sensors Article A Systematic Approach to the Design and Characterization of a Smart Insole for Detecting Vertical Ground Reaction Force (vGRF) in Gait Analysis Anas M. Tahir 1 , Muhammad E. H. Chowdhury 1 , Amith Khandakar 1 , Sara Al-Hamouz 1 , Merna Abdalla 1 , Sara Awadallah 1 , Mamun Bin Ibne Reaz 2 and Nasser Al-Emadi 1, * 1 2 * Department of Electrical Engineering, Qatar University, Doha 2713, Qatar; a.tahir@qu.edu.qa (A.M.T.); mchowdhury@qu.edu.qa (M.E.H.C.); amitk@qu.edu.qa (A.K.); sa1507714@student.qu.edu.qa (S.A.-H.); ma1508307@student.qu.edu.qa (M.A.); sa1509524@student.qu.edu.qa (S.A.) Department of Electrical, Electronic & Systems Engineering, Universiti Kebangsaan Malaysia, Bangi, Selangor 43600, Malaysia; mamun@ukm.edu.my Correspondence: alemadin@qu.edu.qa; Tel.: +974-4403-4213 Received: 7 December 2019; Accepted: 28 January 2020; Published: 11 February 2020



 Abstract: Gait analysis is a systematic study of human locomotion, which can be utilized in various applications, such as rehabilitation, clinical diagnostics and sports activities. The various limitations such as cost, non-portability, long setup time, post-processing time etc., of the current gait analysis techniques have made them unfeasible for individual use. This led to an increase in research interest in developing smart insoles where wearable sensors can be employed to detect vertical ground reaction forces (vGRF) and other gait variables. Smart insoles are flexible, portable and comfortable for gait analysis, and can monitor plantar pressure frequently through embedded sensors that convert the applied pressure to an electrical signal that can be displayed and analyzed further. Several research teams are still working to improve the insoles’ features such as size, sensitivity of insoles sensors, durability, and the intelligence of insoles to monitor and control subjects’ gait by detecting various complications providing recommendation to enhance walking performance. Even though systematic sensor calibration approaches have been followed by different teams to calibrate insoles’ sensor, expensive calibration devices were used for calibration such as universal testing machines or infrared motion capture cameras equipped in motion analysis labs. This paper provides a systematic design and characterization procedure for three different pressure sensors: force-sensitive resistors (FSRs), ceramic piezoelectric sensors, and flexible piezoelectric sensors that can be used for detecting vGRF using a smart insole. A simple calibration method based on a load cell is presented as an alternative to the expensive calibration techniques. In addition, to evaluate the performance of the different sensors as a component for the smart insole, the acquired vGRF from different insoles were used to compare them. The results showed that the FSR is the most effective sensor among the three sensors for smart insole applications, whereas the piezoelectric sensors can be utilized in detecting the start and end of the gait cycle. This study will be useful for any research group in replicating the design of a customized smart insole for gait analysis. Keywords: gait analysis; characterization; smart insole; vertical ground reaction forces; force sensitive resistors; piezoelectric sensors; sensor calibration Sensors 2020, 20, 957; doi:10.3390/s20040957 www.mdpi.com/journal/sensors Sensors 2020, 20, 957 2 of 25 1. Introduction Gait analysis offers an opportunity for assessment of the act of walking, one of the most important features of the individual’s use pattern that displays posture in action. By identifying gait kinetics, gait kinematics and musculoskeletal activity, gait analysis can be utilized in various applications, such as rehabilitation, clinical diagnostics and sport activities [1]. Gait kinetics studies the forces and moments that results in movement of lower extremities during gait cycle. Vertical ground reaction forces (vGRFs) are the forces between the foot and ground which can be obtained by wearable sensors [2] and are considered as the main measurement in kinetic analysis. Gait kinetics have recently become a convenient tool for biomedical research and clinical practice. Different research teams studied the ability to diagnose or early detection of various diseases using gait analysis [3–5]. Some research teams used gait analysis in fall detection of elderly people, one of the most common domestic accidents among the elderly. With smart insoles, the fall event can be detected and doctors or personal who takes care of the elderly can be notified to take action. In athletic sports where walking, running, jumping and throwing are involved, gait analysis can be utilized to recognize an athlete’s faulty movement and, accordingly, enhance it. In addition, gait analysis can play positive role in the rehabilitation process for several diseases and complications. Recently, with the development in sensor technologies, gait analysis using wearable systems became an effective approach [6–8]. Various types of wearable sensors such as force sensors, strain gauges, magneto-resistive sensors, accelerometers, gyroscopes, inclinometers etc. can analyze different gait characteristics. Accelerometers were used to conduct gait analysis studies, in which they were attached to feet or legs to measure the acceleration or velocity of human lateral movements during gait cycles [4]. Gyroscopes were used in gait analysis to measure the changes in orientation of lower body extremes with respect to the vertical axis. Goniometers measured the relative rotational motion between different body segments [2]. Electromagnetic tracking systems were developed as 3D measurement device that can be applied in the kinematic study of body movements [9]. Gait analysis is typically carried out using a force plate system or multi-camera-based system to capture the ground reaction forces (GRF) during different gait cycles. However, this method requires a costly set up and long post-processing time and can measure only limited number of strides. Therefore, it is not affordable by individuals for personal use [3,8,10]. Instrumented trade mills with few force plates laid on the trade mill are used by different research groups to mitigate the limitations of conventional force plates [2], but with treadmills restrictions are still present as subjects need to walk in a straight line where direction changes and turning cannot be realized. This led to an increase in research interest towards developing smart insoles, where wearable sensors can be employed to detect vGRF, joint movements, acceleration of lower extremities, and other gait variables [3,4,11,12]. vGRF is a useful tool to assess the health conditions of the patient, to enhance the performance of athletes [13–15]. Among different solutions for vGRF measurement, smart insoles have several extra advantages over force plates and multi-camera systems. Although force plates can measure shear forces and pressure changes, smart insoles are portable and capable of tracking motions and measuring pressure without rigid mounting, whereas the camera-based system requires large space for set-up along with long post-processing time. The smart insole offers flexible, portable, and comfortable solution for vGRF measurement. It is designed to monitor, process and display plantar pressure using pressure sensors embedded in the insole [3,4,11,12]. Recently, several off-the-shelf smart insoles have been offered by some companies (e.g., F-scan [16], MoveSole [17], Bonbouton [18], FeetMe [19] etc.), however, the commercial systems are very expensive for individual use, making it difficult for a home setting. The aim of this study is to design and characterize smart insoles to detect vGRF during gait, with three different types of low-cost commercial force sensor: force-sensitive resistors (FSRs) [20], ceramic piezoelectric sensors [21], and flexible piezoelectric sensors [22]. All three types of sensor were calibrated before checking their suitability for smart insole application. A simple low-cost calibration method based on load cells is presented, mitigating the need to use expensive calibration devices or Motion Analysis Labs as a calibration reference. This work provides a systematic approach for sensor Sensors 2020, 20, 957 3 of 25 calibration guides, which can be replicated easily by other researchers to perform studies on smart insoles or other body-sensing technologies. To the best of our knowledge, this is the first article to compare three different low-cost commercially available force sensors for smart insole application. The remainder of the article is organized into five sections. In Section 2, a comprehensive review of the recent works with smart insoles to detect vGRF in gait cycles are summarized. In Section 3, the experimental details for sensors calibration and insole characterization are presented. In Section 4, the mathematical analysis of each insole characterization and sensor calibration are explained. Results and a discussion are presented in Section 5. Finally, we conclude with future recommendations in Section 6. 2. Literature Review Several research teams focused on fabricating and synthesizing the sensing parts or sensing fabrics of the smart insoles [23–25]. Sensing fabrics are fibers/yarns with sensing technologies or electrical components made of fabric materials, offering a flexible alternative to comfortably measuring human movement. Usually, piezoelectric, piezoresistive and piezo-capacitive materials are used to fabricate the sensing parts of the sensing fabrics, due to their elastic properties [26,27]. Shu et al. [26] implemented a low-cost insole with high pressure sensitivity using a fabric pressure sensing array made by the researchers with a pressure range of 10 Pa to 1000 kPa. It is attached to six locations corresponding to a polyimide film circuit board that takes the shape of the foot. They were able to measure the peak pressure, mean pressure, center of pressure (COP), and illustrate different pressure levels occurring at the six-targeted areas. However, the quality of the gait cycle records was poor, with irregular peak values, where the common gait shape with two peaks of the heel strike and toe off cannot be distinguished. Kessler et al. [27] demonstrated a low-cost flexible insole, made with Velostat and conductive ink electrodes printed on polyethylene terephthalate (PET) substrate. However, repeatability was a major problem and they proposed an averaging method to reduce the repeatability issue. However, the proposed method does not provide a generic solution for the force-sensing problem, it can be utilized only with periodic forces where spatial information is the key. On the other hand, some research teams used low-cost flexible force sensors to design the smart insoles [28–30] using commercially available piezoresistive [20], piezoelectric [21,22], capacitive transducers [2], fiber brag grating [5,31] sensors. Piezoelectric force sensors are materials that generate electric charges when stressed. However, there are a few factors which limit the usage of piezoelectric sensors in smart insoles. The parasitic effect of piezo materials neutralizes the generated charge within a short time. Therefore, sophisticated electronics are needed to extract resultant charges, and this makes it difficult to use these sensors in measuring static or slow varying forces. In addition, protection circuits are needed, since piezo sensors generate high voltage values, which might reach above 100 V with peak vGRF values. Capacitive force sensors are another alternative force sensor, consisting of parallel capacitor plates that changes the capacitance in correspondence to applied force/weight. However, they need complex conditioning circuits and are highly subject to noise [20]. A commonly used body-sensing technology is the piezo-resistive sensor or FSR, which changes its conductivity based on the applied force. FSR is a polymer thick film (PTF) that is used to measure the applied force in different applications such as human touch and medical applications, industrial and robotics applications, and automotive electronics. The main advantages of FSRs are: thin size, very good shock resistance, low power requirement, fast response to force changes, robustness against noise, simple conditioning circuits, ability to fabricate using flexible materials, and low unit cost compared to other commercial force sensors [20]. However, these sensors have some disadvantages that need to be compensated for, such as non-linear behavior and repeatability error [3]. Bamberg et al. [4] used a combination of different FSRs, piezo electric sensors, accelerometers and gyroscopes to determine the vGRF. The main advantage of this approach is that it enables the detection of heel strike and toe off events in each gait cycle. In addition, it helps in estimating foot orientation and position. Even though gait variability can be analyzed by walking in a straight line, Sensors 2020, 20, 957 4 of 25 gait analysis concentrating merely on straight walking or running may not be adequate to interpret gait variability, since changing walking directions or turning have effects on extrinsic gait variability [11]. Similar research was done recently in [32], where the research group used the FSR sensor to develop the smart foot sole which transmits wirelessly the vGRF to a computer, and the patients were asked to walk on treadmill during the signal acquisition. Liu et al. [11] developed a wearable measuring insole using five triaxial force sensors in each shoe capable of measuring GRF and center of pressure (COP) on insole. The GRF results showed a great correspondence between the insole and the reference data. Kim et al. [33] conducted a similar study, where they have used similar triaxial force sensors and the sensors performance were tested on seven healthy male subjects. An in-shoe plantar measurement sensor with 64 sensing points made from an optoelectronics transducer covered with silicon in a matrix form covering 80% of contact region between the foot and the insole and handling capability of 1MPa was implemented by De Rossi et al. to measure COP and vGRF [5]. Howel et al. [3] demonstrated the design of a wearable smart insole using low-cost FSRs for gait analysis. This provided subject-specific linear regression models to determine the vGRF accurately using simultaneous collected data from motion analysis laboratory. However, insufficient information was given about the sensors calibration and the hardware design of the insole and the wireless system to transmit the data to host PC, making it difficult for other researchers to replicate the work. Even though systematic sensor calibration with clear steps was followed by different research teams, expensive calibration devices were used to calibrate the force sensors. Some research teams carried out the experiments on the smart insoles in motion analysis labs, where simultaneous data collection from infrared motion capture cameras/RGB depth camera and force plates were done as reference measurement for the collected insole data [34,35]. In addition, some research teams used a universal testing machine to apply incremental weight values to sensor active area during calibration. Barnea et al. [36] used the CETR Universal Micro-Tribometer (UMT)-2 micro tribometer) device for calibrations, that can apply precise weights in X, Y and Z directions. Marco et al. [5] performed the sensor calibrations using robotic platform that can precisely apply controllable loads to the desired positions. Parmar et al. [37] evaluated the performance of 5 different commercial FSRs during static and dynamic loading with reliable test setups that can mimic realistic conditions when applying pressure on human limbs. The sensors were evaluated quantitatively based on their accuracy, drift, and repeatability behaviors. The tested sensors showed lower accuracy levels with static pressures compared to the dynamic pressure test, with high drift values. This necessitates the need for further study and analysis on the use of FSRs for static pressure applications. 3. Methodology This section demonstrates the design of a complete system describing the main blocks of the smart insole along with illustrations of sensor calibration and insole characterization process. 3.1. Smart Insole Sub-System Figure 1 shows the complete block diagram of the system, where the pressure sensor array was placed in a customized shoe above the control circuit. Pressure data were digitized through a microcontroller before they were sent wirelessly to a host computer for post processing and analysis. This subsystem was powered by a battery with the help of a power management unit. Pressure data were analyzed to extract various gait characteristics for different gait applications. Sensors 2020, 20, 957 Sensors 2020, 20, 957 5 of 25 5 of 25 relation between the applied force and the sensor’s resistance. In addition, no direct relationship is relation between the applied force and the sensor’s resistance. In addition, no direct relationship is 5 of 25 sensor’s datasheet. Therefore, proper calibration must be done prior to the sensor provided in the sensor’s datasheet. Therefore, proper calibration must be done prior to the sensor usage. usage. Sensors 2020, 20, provided in 957 the Figure1.1.Smart Smartinsole insoleblock blockdiagram. diagram. Figure Figure 1. Smart insole block diagram. 3.1.1. B. Pressure-Sensing Array Sensor Ceramic Piezoelectric B. Ceramic Piezoelectric Sensor A piezoelectric element is a sensor that produces an alternating voltage in response to an applied The vGRF during gait cycles can bethat sensed usingan one of three alternatives: A piezoelectric element is a sensor produces alternating voltage in response to an applied dynamic pressure or vibration. With applications related to dynamic forces, the piezoelectric sensor A. Force-Sensitive Resistor (FSR) dynamic pressure or vibration. With applications related to dynamic forces, the piezoelectric sensor is highly recommended. When a force applied to the piezoelectric crystal element, the net movement The FSR exhibits a decrease resistance asto the force tocrystal the surface of the increases. is highly recommended. When ain force applied theapplied piezoelectric element, thesensor net movement of both positive and negative ions occurs. When there is a constant or zero pressure, the dipole is not FSR sensors from Interlink Electronics [20] were used in this study as shown in Figure 2A. The sensors of both positive and negative ions occurs. When there is a constant or zero pressure, the dipole is not formed [38]. Itround is important to mention that the force plate is originally made of piezoelectric material have a flexible activeto area of diameter 12.7 mm to detect the applied with a twomaterial flexible formed [38]. It is important mention that the force plate is originally madeforce, of piezoelectric mounted between two metal plates to produce three‐dimensional forces with a special mechanical lead wires to connect the sensor to the acquisition circuit. A FSR exhibits a non-linear relation between mounted between two metal plates to produce three‐dimensional forces with a special mechanical arrangement [39]. This comes in different sizes; however, ceramic piezoelectriciselement with 12.8 the applied force sensor’s resistance. Inhowever, addition,aano direct piezoelectric relationship provided in12.8 the arrangement [39].and Thisthe comes in different sizes; ceramic element with mm electrode diameter would be suitable to obtain a high‐resolution pressure map as shown in sensor’s datasheet. Therefore, calibration must abehigh‐resolution done prior to the sensor usage. mm electrode diameter wouldproper be suitable to obtain pressure map as shown in Figure 2b. B. 2b.Ceramic Piezoelectric Sensor Figure C. Micro‐Electromechanical Systems (MEMS) Sensor A element is a sensor that (MEMS) produces an alternating voltage in response to an applied C. piezoelectric Micro‐Electromechanical Systems Sensor The micro‐electromechanical systems (MEMS) sensor is a newforces, member of piezoelectric sensors dynamic pressure or vibration. With applications related to dynamic the is The micro‐electromechanical systems (MEMS) sensor is a new member ofpiezoelectric piezoelectricsensor sensors family (Figure 2c). Similar to ceramic piezo electric sensors, it converts mechanical forces into highly recommended. When a force applied to the piezoelectric crystal element, the net movement of family (Figure 2c). Similar to ceramic piezo electric sensors, it converts mechanical forces into electrical signals. However, the occurs. MEMS sensor can detect forces in x, zero y or z axes generating electrical both positive andHowever, negative ions Whencan there is a constant the dipole is not electrical signals. the MEMS sensor detect forces in or x, y or zpressure, axes generating electrical impulses with positive or negative amplitudes depending on the force direction on a certain axis[40]. formed [38]. It is important to mention that the force plate is originally made of piezoelectric material impulses with positive or negative amplitudes depending on the force direction on a certain axis[40]. MEMS sensors are useful for plates detecting human motion sensor due to theirwith flexibility, wide frequency mounted between metal to human produce three-dimensional a special MEMS sensors are two useful for detecting motion sensor due toforces their flexibility, widemechanical frequency range (0.001 Hz to 10 MHz), low acoustic impedance, high mechanical strengths, and high12.8 stability arrangement [39]. This comes in different sizes; however, a ceramic piezoelectric element with mm range (0.001 Hz to 10 MHz), low acoustic impedance, high mechanical strengths, and high stability resisting moisture, etc. [40]. electrode diameter would be suitable to obtain a high-resolution pressure map as shown in Figure 2B. resisting moisture, etc. [40]. A A B B C C Figure 2. (A) Force‐sensitive resistor (FSR) sensor from Interlink Electronics [20], (B) piezo‐electric piezo‐electric Figure 2. (A) Force‐sensitive Force-sensitive resistor resistor (FSR) (FSR) sensor sensor from from Interlink Interlink Electronics Electronics [20] [20],, (B) piezo-electric sensor from Murata Manufacturing Co.[38], (C) micro‐electromechanical systems (MEMS) sensor sensor from Murata [38],, (C) micro-electromechanical Murata Manufacturing Manufacturing Co. [38] micro‐electromechanical systems (MEMS) sensor LDT0‐028K from Measurement Specialties Inc. [40]. LDT0-028K LDT0‐028K from Measurement Specialties Specialties Inc. Inc. [40]. [40]. C. Micro-Electromechanical Systems (MEMS) Sensor Sensors 2020, 20, 957 6 of 25 The micro-electromechanical systems (MEMS) sensor is a new member of piezoelectric sensors family (Figure 2C). Similar to ceramic piezo electric sensors, it converts mechanical forces into electrical signals. However, the MEMS sensor can detect forces in x, y or z axes generating electrical impulses with positive or negative amplitudes depending on the force direction on a certain axis [40]. MEMS sensors are useful for detecting human motion sensor due to their flexibility, wide frequency range (0.001 Hz to 10 MHz), low acoustic impedance, high mechanical strengths, and high stability resisting moisture, etc. [40]. 3.1.2. Data Acquisition System A. Microcontroller (MCU): A microcontroller (MCU) was used to collect the data from the sensor and to send to the computer for classification. Simblee is a very compact and powerful ARM Cortex-M0 MCU with a six channels 10-bit analog-to-digital converter (ADC). It is featured with an inter-integrated circuit (I2 C) and serial peripheral interface (SPI) communication interface, which were required for 9-degree of freedom (DOF) module. Moreover, it has an incorporated Bluetooth low energy (BLE) 4.0 module, which can be utilized to send data to the computer. This MCU operates on a power supply between +2.1 to 3.6 V. B. Multiplexer (MUX) Since MCU has a limited number of ADC channels whereas the number of sensors is needed for better spatial resolution of smart insole, it is suggested to use multiplexers (MUX) to reduce the number of required channels in MCU. A MUX allows several inputs in parallel to be routed into a single output depending on the input combinations of the data selectors. Active area of these sensors are close and sixteen sensors were used to create sensors’ array for each leg insole to obtain a high-resolution pressure map. Therefore, the CD74HC4067 multiplexer from Texas Instruments with 16 input channels was used in this study [41]. 3.1.3. Transmission Techniques Three commonly used transmission techniques for connected biomedical sensors are ZigBee, Bluetooth Low Energy (BLE) and Wi-Fi. ZigBee is a two-way wireless communication technique developed for sensors and control networks, which need a wider range, low latency, low energy consumption at lower data rates. BLE is an alternative to the classical Bluetooth with higher data rate and low power consumption within a limited area with low latency at 2.4 GHz. Wi-Fi makes a good candidate for transmitting data with a data rate of up to 450 Mbps for indoor applications. However, it imposes latency on the system of more than 25 ms and higher power consumption. Table 1 shows a comparison between three different communication interfaces. Table 1. Transmission methods comparison [42–46]. Latency Speed Power Consumption Range ZigBee Bluetooth Low Energy 15 ms 250 Kbps 9.3 mA 291 m 6 ms 1–11 Mbps 4.5 mA 10 m Wi-Fi ≥25 ms 1.3 Gbps over 5 GHz and 450 Mbps over 2.4 GHz 35 mA 50 m Since the smart insole was intended for indoor application, BLE and WiFi both were suitable for communication interface; however, the higher power consumption and latency made WiFi non-suitable for smart insole application. Moreover, Simblee MCU has in-built BLE in its small form factor. Therefore, BLE has been chosen as communication interface. Sensors 2020, 20, 957 7 of 25 3.1.4. Power Management Unit (PMU) Power supplies were chosen depending on the operating voltage of the system components. The microcontroller and multiplexer both can operate at 3.3 V. The power management unit (PMU) is LiPo Charger/Booster module MCP73831 [47] and AMS1117 voltage regulator connected to a Lithium Polymer (LiPo) battery of 3.7 V (1000 mAh), which was regulated to 3.3 V. The PMU is not only delivering regulated 3.3 V to the system but also capable to charge LiPo battery. 3.1.5. Host Computer The acquired data from smart insole can be sent wirelessly to a host computer, where post processing, thereby displaying the vGRF as pressure maps during gait cycle, was carried out. The obtained data can be used in different gait analysis applications such as medical diagnostics, rehabilitation and athlete’s performance assessment. 3.2. Sensors’ Calibration The first step in designing the smart insole is to calibrate the force sensors that are going to be used to detect the vGRF during the gait cycle. Three different force sensors were calibrated: FSR [20], piezo-electric sensor [21] and piezo -vibration sensor [22]. 3.2.1. Force-Sensitive Resistor (FSR) Calibration Firstly, a voltage divider circuit must be used with the sensor to convert the resistance change (due to applied force) of the sensor to a voltage value, which can be acquired by microcontrollers. Secondly, a load cell of 5kg from HT sensor technology company [48], with HX711 amplifier modules [15] was used as a weight reference for FSR calibration (Figure 3). The load cells consist of straight metal bar with two strain gauge sensors and two normal resistors arranged in a Whitestone bridge configuration, a constant excitation voltage (3–5 V) can be applied as an input to the circuit and the balanced configuration of the circuit replicates a zero output voltage in normal conditions when no force is applied. Any force applied to the load cell results in an unbalanced condition of the bridge leading to small voltage values in the output that can be detected and converted to force [48]. The load cell has high sensitivity and can detect as small as 1 gm of weight variation. However, the output voltage from the load cell is very small, with a maximum value of 5 mV. Therefore, a HX711 amplifier module was used. The amplifier module has instrumentational amplifier to amplify the signal with a 24-bit ADC that converts the analog signal from the load cell bridge to digital value that is readable by a microcontroller. The HX711 transmits data to the microcontroller using I2 C communication protocol with 10Hz sampling rate [15]. The bar-type load cell was mounted with screws and spacers so that the strain can be measured correctly (refer Figure 3B). The load cell was placed between two plates with only one side screwed into each plate/board. This setup provides a moment of force on the strain gauges rather than just a single compression force, resulting in higher sensitivity to applied forces. The output voltage from the load cell exhibits a linear relationship with the applied force. This can be calibrated easily with any small object of known mass such as a coin that weighs a few grams. A known weight object (ex. a coin) was placed on load cell plate; the calibration factor was adjusted until the output reading matches the known weight. Once the correct calibration factor is obtained, it was used to convert the load cell voltages to corresponding weights. The calibration factor is the slope of output voltage of load cell vs. real weights’ graph. The FSR was attached to adhesive material on the back face of the active area, which was used to fix the FSR on the scale. A cylindrical acrylic of 12.7 mm diameter, matching the active area of FSR, was used to apply force on the sensor only. In addition, a square shaped acrylic plate was glued on top of the cylindrical acrylic to support the weights, as shown in Figure 3. Then, 500 g weights are placed every 4 to 5 s until 5000 g is reached. Readings from load cell and FSR circuit are acquired simultaneously by Arduino, which were saved in a text file in a computer. configuration, a constant excitation voltage (3–5 V) can be applied as an input to the circuit and the balanced configuration of the circuit replicates a zero output voltage in normal conditions when no force is applied. Any force applied to the load cell results in an unbalanced condition of the bridge leading to small voltage values in the output that can be detected and converted to force [48]. The load cell Sensors 2020,has 20, high 957 sensitivity and can detect as small as 1 gm of weight variation. However, the output 8 of 25 voltage from the load cell is very small, with a maximum value of 5 mV. Therefore, a HX711 amplifier module was used. The amplifier module has instrumentational amplifier to amplify the signal with Finally, the FSR output voltage wassignal plotted withthe respect to the load-cell weight and athat mathematical a 24‐bit ADC that converts the analog from load cell bridge to digital value is readable relationship was derived. The equation was used to convert smarts insole FSR readings into the 2 by a microcontroller. The HX711 transmits data to the microcontroller using I C communication corresponding applied pressure by[15]. the foot. protocol with 10Hz sampling rate A Sensors 2020, 20, 957 8 of 25 B Figure 3. FSR calibration setup and load‐cell scale Figure 3. FSR calibration setup (A)(a) and load-cell scale (B).(b). 3.2.2. Piezo-Electric Calibration The bar‐typeSensor load cell was mounted with screws and spacers so that the strain can be measured correctly (refer Figuremodule 3b). The load cellfor was placed betweenwith twosome plates with only one screwed The same load-cell was used piezo calibration modifications (asside shown in into each plate/board. This setup provides a moment of force on the strain gauges rather than just Figure 4). Piezo transducers convert the applied mechanical forces into electrical impulses. Therefore, a single compression force,(above resulting in higher sensitivity to applied forces. output voltage from the a high sampling frequency 50 Hz) is needed to acquire both the piezoThe output and the applied load cell exhibits a linear relationship with the applied force. This can be calibrated easily with weights from the load cells. HX711 amplifier module samples the data with a low sampling frequencyany small of known mass such as a coin that weighs a few ADC grams. of 10 Hz.object Therefore, the data was acquired directly by the 10-bit of Arduino MCU with a sampling A known weight object AD620AN (ex. a coin)instrumentational was placed on load cell plate; calibration factorthe was frequency of 1 kHz. However, amplifiers [39]the were used before adjusted until the output reading matches the known weight. Once the correct calibration factor is acquisition step to amplify the small load cell outputs (maximum of 5 mV). obtained, it was used to convert the load cell voltages to corresponding weights. The calibration factor Firstly, a voltage divider circuit was used to reduce the high piezo voltage outputs, which can go slope of output loadcalibrated cell vs. real weights’ The FSR was attached adhesive upistothe 20 V. Secondly, the voltage load cellofwas again due tograph. the modification. Three deadto weights material on the back face of the active area, which was used to fix the FSR on the scale. A cylindrical of known masses were used: 500 kg, 2500 kg 5000 kg (maximum load for load cell). AD620AN acrylic of 12.7 mm diameter, matching the active area FSR, was used towhen applymaximum force on the sensor instrumentational amplifier gain was adjusted to give anofoutput of voltage load is only. In addition, a square shaped acrylic plate was glued on top of the cylindrical acrylic to support applied. This ensures that the full range of the Arduino ADC was utilized. Three dead weights were the weights, as shown Figure Then, 500 gvoltage weights are placed every 4 toacquired 5 s until by 5000 g is reached. added one by one on thein scale and3.the output from load cells were Arduino. A Readings from load andbetween FSR circuit acquired simultaneously by Arduino, were saved in linear relationship wascell fitted theare load cell voltages and applied weights. which This relationship a text a computer. was usedfile to in convert the load cell voltage to a corresponding weight. Finally, the FSR output voltage wastoplotted with to thesince load‐cell Unlike the FSR, weights cannot be used calibrate the respect piezo sensor, piezo weight sensors and are a mathematical relationship wasTherefore, derived. The was and usedrelease to convert smarts insole FSR readings sensitive to dynamic forces only. a fastequation finger press is suggested as an alternative. into the corresponding applied pressure by the foot. The calibration can be done by pressing the active area/ceramic of the sensor with various strengths 3.2.2. Piezo‐Electric Sensor Calibration The same load‐cell module was used for piezo calibration with some modifications (as shown in Figure 4). Piezo transducers convert the applied mechanical forces into electrical impulses. Therefore, a high sampling frequency (above 50 Hz) is needed to acquire both the piezo output and the applied Sensors 2020, 20, 957 9 of 25 and recording the generated electrical signals for each press as shown in Figure 4. Readings from the load cell and piezo voltage divider circuit were acquired simultaneously by the Arduino MCU. Serial terminal software was used to store the data in the computer. Load cell readings were plotted against the piezo output voltage and a linear relationship was derived. The equation was used to convert smarts insole readings into the corresponding applied weight by the foot. Sensors 2020, 20, 957 9 of 25 Figure Figure 4. 4. Piezoelectric Piezoelectric sensor sensor calibration calibration setup. setup. 3.2.3. Micro-Electromechanical Systems (MEMS) Sensor Calibration Firstly, a voltage divider circuit was used to reduce the high piezo voltage outputs, which can MEMS producethe alternating (AC) impulses and negative go up to 20sensors V. Secondly, load cell current was calibrated again with due both to thepositive modification. Threepeaks. dead Therefore, modification was required setup piezo vibration refer to weights oflittle known masses were used: 500for kg,the2500 kgof5000 kgelectric (maximum loadsensor, for load cell). Figure 5. An offset circuit was added for the piezo-electric acquisition circuit. The piezo-vibration AD620AN instrumentational amplifier gain was adjusted to give an output of voltage when output voltage reducedThis by aensures voltagethat divider circuit to ±1/2 wasThree used maximum loadwas is applied. the full range of theVArduino ADC amplifier was utilized. cc , then adder to add an offset of +1/2 V , so the new AC signal will be centered around +1/2 V with maximum dead weights were added cc one by one on the scale and the output voltage from cc load cells were value of Vby minimum of 0 V. After modifying the acquisition circuit, electric acquired Arduino. A linear relationship was fitted between the load the cell piezo voltages and sensors’ applied cc and calibration steps were used to calibrate the piezo-vibration sensor. weights. This relationship was used to convert the load cell voltage to a corresponding weight. Sensors 2020, 20, 957 10 of 25 Unlike the FSR, weights cannot be used to calibrate the piezo sensor, since piezo sensors are sensitive to dynamic forces only. Therefore, a fast finger press and release is suggested as an alternative. The calibration can be done by pressing the active area/ceramic of the sensor with various strengths and recording the generated electrical signals for each press as shown in Figure 4. Readings from the load cell and piezo voltage divider circuit were acquired simultaneously by the Arduino MCU. Serial terminal software was used to store the data in the computer. Load cell readings were plotted against the piezo output voltage and a linear relationship was derived. The equation was used to convert smarts insole readings into the corresponding applied weight by the foot. 3.2.3. Micro‐Electromechanical Systems (MEMS) Sensor Calibration MEMS sensors produce alternating current (AC) impulses with both positive and negative peaks. Therefore, little modification was required for the setup of piezo electric vibration sensor, refer to Figure 5. An offset circuit was added for the piezo‐electric acquisition circuit. The piezo‐vibration output voltage was reduced by a voltage divider circuit to ±1/2 Vcc, then adder amplifier was used to add an offset of +1/2 Vcc, so the new AC signal will be centered around +1/2 Vcc with maximum value of Vcc and minimum of 0 V. After modifying the acquisition circuit, the piezo electric sensors’ Figure 5. MEMS setup. 5. LDT0-028k LDT0‐028k MEMS sensor sensor calibration calibration calibration steps were usedFigure to calibrate the piezo‐vibration sensor. setup. 3.3. Insole Fabrication Once the sensors were calibrated, these sensors were separately used to construct the smart insole for vGRF detection during gait cycles. The FSR sensors and piezo‐electric sensors were chosen to construct two different insoles. While the piezo‐vibration sensor was found not suitable for vGRF detection, the reason of not selecting the piezo‐vibration sensor is discussed in a later section. As shown in Figure 6, the most common place of the foot plane, where most of the pressure is exerted during gait are the heel, metatarsal heads, hallux and toe. Sixteen sensors were placed on each insole to record pressure values in these areas. While no sensors were placed on the medial arch area of the foot as most people exert very low/no pressure on Sensors 2020,due 20, 957 10 of 25 that area to it is arch shape [49]. Smart insole data were collected from 16 FSRs/piezo‐electric sensors. Sixteen inputs were multiplexed to one output through a 16‐to‐1 multiplexer and applied to an ADC input of the microcontroller then sent to host computer. All subjects were asked to place the 3.3. Insole Fabrication sensor’s insole inside their shoes, then placing another layer of insole on top of it to ensure comfort of the subject while walking. The acquisition and were transmission were connected through a Once the sensors were calibrated, these sensors separatelycircuit used to construct the smart insole conductive pathwayduring that can in minimizing the size of piezo-electric the wire and sensors avoidingwere anychosen electrical for vGRF detection gaithelp cycles. The FSR sensors and to hazard. were worn While by thethe subject inside his/her ownwas shoe while acquisition and constructThe twoinsoles different insoles. piezo-vibration sensor found notthe suitable for vGRF transmission placed inside × 7cm box attached the subject’s by an adhesive detection, thecircuits reasonwere of not selecting thea 7cm piezo-vibration sensor istodiscussed in aleg later section. As strap belt thecommon data. Acquired data sent viawhere Bluetooth computer, they shown in while Figureacquiring 6, the most place of thewere foot plane, mostto ofathe pressurewhere is exerted were plotted and during gait are theanalyzed. heel, metatarsal heads, hallux and toe. Figure 6. Area Area of of foot foot selected selected for sensors (A), and array of pressure sensor (B) in those areas. sensors were placed on each insole to record pressure values in these areas. While no 3.3.1.Sixteen FSR Insole Characterization sensors were placed on the medial arch area of the foot as most people exert very low/no pressure on that area due to it is arch shape [49]. Smart insole data were collected from 16 FSRs/piezo-electric sensors. Sixteen inputs were multiplexed to one output through a 16-to-1 multiplexer and applied to an ADC input of the microcontroller then sent to host computer. All subjects were asked to place the sensor’s insole inside their shoes, then placing another layer of insole on top of it to ensure comfort of the subject while walking. The acquisition and transmission circuit were connected through a conductive pathway that can help in minimizing the size of the wire and avoiding any electrical hazard. The insoles were worn by the subject inside his/her own shoe while the acquisition and transmission circuits were placed inside a 7cm × 7cm box attached to the subject’s leg by an adhesive strap belt while acquiring the data. Acquired data were sent via Bluetooth to a computer, where they were plotted and analyzed. 3.3.1. FSR Insole Characterization Twelve healthy subjects (Table 2) were asked to walk a straight 10 m walkway with self-selected cadence six times with an average walking speed of 3–4 mile per hour (MPH) and data acquired at 60 Hz sampling frequency using the smart insole made up of 16 FSRs (Figure 7). On treadmills, participants are restricted to walk in straight line as direction changes and turning cannot be realized; however, in the proposed study, the user walked freely in a 10-m walkway and they were asked to walk in a corridor which has a length of 10m and width of 1.5 m and they did not need to walk completely in a straight path and the user can walk in self-cadence, which is not possible on a treadmill. Subjects were asked to place the smart insole in their shoe while wearing cotton socks to avoid any sweat leakage that might damage the sensors or affect data acquisition from the sensors. Although walking speed is an important factor in some applications, it is not needed in many gait studies where the main focus is to detect the vertical ground reaction forces and asses the gait variables. The statistical gait variables were the symmetry between both feet, percentage of different phases (stance and swing phase) and sub phases (heel strike, mid-stance, toe off etc.) in a full gait cycle. Those statistical variables were used in various studies including sports or medical applications for gait analysis, without the need for participants are restricted to walk in straight line as direction changes and turning cannot be realized; walk in a corridor which has a length of 10m and width of 1.5 m and they did not need to walk however, in the proposed study, the user walked freely in a 10‐m walkway and they were asked to completely in a straight path and the user can walk in self‐cadence, which is not possible on a walk in a corridor which has a length of 10m and width of 1.5 m and they did not need to walk treadmill. Subjects were asked to place the smart insole in their shoe while wearing cotton socks to completely in a straight path and the user can walk in self‐cadence, which is not possible on a avoid any sweat leakage that might damage the sensors or affect data acquisition from the sensors. treadmill. Subjects were asked to place the smart insole in their shoe while wearing cotton socks to Although walking speed is an important factor in some applications, it is not needed in many gait avoid any sweat leakage that might damage the sensors or affect data acquisition from the sensors. 11 of 25 Sensors 2020, 20, 957 studies where the main focus is to detect the vertical ground reaction forces and asses the gait Although walking speed is an important factor in some applications, it is not needed in many gait variables. The statistical gait variables were the symmetry between both feet, percentage of different studies where the main focus is to detect the vertical ground reaction forces and asses the gait phases (stance and swing phase) and sub phases (heel strike, mid‐stance, toe off etc.) in a full gait variables. The statistical gait variables were the symmetry between both feet, percentage of different walkingcycle. speed measurement. However, the walking recorded the impact of walking Those statistical variables were used in various speed studieswas including sportsto orsee medical applications phases (stance and swing phase) and sub phases (heel strike, mid‐stance, toe off etc.) in a full gait speed inforthe forwithout a gait cycle. The data were converted into forcethe values byspeed the relationship gaitvGRF analysis, the need for FSR walking speed measurement. However, walking was cycle. Those statistical variables were used in various studies including sports or medical applications recorded to see the impact of walking speed in thedata vGRFwere for a gait cycle. FSR datainstance were converted obtained theanalysis, calibration stage. Then sensors’ added atThe each time towas obtain one forin gait without the need for16 walking speed measurement. However, the walking speed into force values by the relationship obtained in the calibration stage. Then 16 sensors’ data were value that represents theimpact full force exerted byinthe while walking (i.e., recorded to see the of walking speed thebody vGRF for a gait cycle. The FSRvGRF). data were converted added at each time instance to obtain one value that represents the full force exerted by the body into force values by the relationship obtained in the calibration stage. Then 16 sensors’ data were while walking (i.e., vGRF). added at each time instance to 2. obtain one value variables that represents the full force exerted by the body Table Demographic of participants. while walking (i.e., vGRF). Table 2. Demographic variables of participants. Body Mass Number of Gender Age (Year) Weight (kg) Height (cm) 2 Index (kg/m Subjects Table 2. Demographic variables of participants. Body Mass Index) Gender Number of Subjects Age (Year) Weight (kg) Height (cm) 2) 7 30.1 ± 13.1 77.3 ± 21.2 159.8 ± 4.9 Body (kg/m 30.3 7.9 Female Mass±Index Gender Number of Subjects Age (Year) Weight (kg) Height (cm) 7 30.1 ± 13.1 77.3 ± 21.2 159.8 ± 4.9 30.3 ± 7.9 FemaleMale 5 52.3 ± 4.8 83.5 ± 3.34 172.7 ± 11.7 28.3 ±2) 2.9 (kg/m 5 52.3 ± 4.8 83.5 ± 3.34 172.7 ± 11.7 28.3 ± 2.9 Male 7 30.1 ± 13.1 77.3 ± 21.2 159.8 ± 4.9 30.3 ± 7.9 Female 5 52.3 ± 4.8 83.5 ± 3.34 172.7 ± 11.7 28.3 ± 2.9 Male Figure 7. Smart Insole using FSR sensor: top (A) and bottom (B). Figure 7. Smart Insole using FSR sensor: top (A) and bottom (B). Figure 7. Smart Insole using FSR sensor: top (A) and bottom (B). 3.3.2. Piezo‐Electric Insole Characterization 3.3.2. Piezo-Electric Insole Characterization 3.3.2.A Piezo‐Electric Insole Characterization similar test was carried out with the piezo‐electric sensor based smart insoles (Figure 8). Three A similar test was carried out with the piezo-electric sensor based smart insoles (Figure 8). Three subjects were asked to walk in a 10 m walkway in normal cadence, with three trials carried out by A similar test was carried the piezo‐electric sensor based smart insoles (Figure 8). Three subjectseach were asked to walk in a acquired 10out mwith walkway in normal cadence, with subject. The data were with a sampling frequency of 60 Hz. three trials carried out by each subjects were asked to walk in a 10 m walkway in normal cadence, with three trials carried out by subject. The data were acquired with a sampling frequency of 60 Hz. each subject. The data were acquired with a sampling frequency of 60 Hz. Figure 8. (A) Piezo insole with 16 piezo sensors, (B) additional insole layer placed on top on piezo insole to ensure comfortability. (A) Piezo insole with piezosensors, sensors, (B) insole layerlayer placed on topon on top piezo Figure Figure 8. (A) 8.Piezo insole with 16 16 piezo (B)additional additional insole placed on piezo insole to ensure comfortability. insole to ensure comfortability. 3.4. Performance Evaluation of the Prototype System A commercial F-scan smart insole system (Figure 9A) was used to validate the designed insole. The F-scan system is one of the best insoles currently available on the market. The insole comes with ultra-thin (0.18 mm) flexible printed circuit with 960 sensing nodes. Each sensing element was recorded with 8-bit resolution with a scanning speed up to 750 Hz. However, the overall cost of the system is 13,000 $ for the wired system and 17,000 $ for the wireless system. On the other hand, the instrumented insole costs only ~500 $. Usually, the vGRF peak is around ±10% of the subject’s weight. Therefore, using F-scan software, data collected from each subject was calibrated based on subject’s weight. The user needs to stand on one foot applying his/her full weight on the insole for 4 to 5 s, then the average applied weight was calculated. If the value obtained was less than the subject’s weight, the F-scan software adjusted the output by a multiplication factor. Similar approach was used in the prototyped FSR insole as well. Figure 9B,C show the F-scan and prototyped system worn by the same subject to compare the vGRF signal acquired by the individual system. instrumented insole costs only ~500 $. Usually, the vGRF peak is around ±10% of the subject’s weight. Therefore, using F‐scan software, data collected from each subject was calibrated based on subject’s weight. The user needs to stand on one foot applying his/her full weight on the insole for 4 to 5 s, then the average applied weight was calculated. If the value obtained was less than the subject’s weight, the F‐scan software adjusted the output by a multiplication factor. Similar approach was used Sensors 2020, 20, 957 in the prototyped FSR insole as well. Figure 9B,C show the F‐scan and prototyped system worn by 12 of 25 the same subject to compare the vGRF signal acquired by the individual system. F‐scan commercial system (A), F‐scan system worn by Subject 01 (B) and FSR‐based Figure 9.Figure F-scan9.commercial system (A), F-scan system worn by Subject 01 (B) and FSR-based prototype prototype system worn by Subject 01 (C). system worn by Subject 01 (C). 4. Analysis 4. Analysis This section explains the mathematical calculations and analyses used for the sensor calibrations Thisand section the mathematical calculations and analyses used for the sensor calibrations insole explains characterization. and insole characterization. 4.1. Sensors’ Calibration 4.1. Sensors’ Calibration 4.1.1. FSR Sensor Calibration 4.1.1. FSR Sensor Calibration The FSR sensors exhibits resistance change in correspondence to the applied force. Therefore, a divider circuit was used to convert the resistance changes to voltage values to be acquired by Thevoltage FSR sensors exhibits resistance change in correspondence to the applied force. Therefore, a microcontroller. voltage divider circuit was used to convert the resistance changes to voltage values to be acquired 𝑅 11 𝑘𝛺 by microcontroller. (1) 𝑉 𝑉 5𝑉 𝑅 𝐹𝑆𝑅 11 𝑘𝛺 𝐹𝑆𝑅 R 11 kΩ Vout = VCCthe × FSR resistance = 5V × decreases, showing an increased output (1) As the applied force increases, also R + FSR 11 kΩ + FSR voltage according to Equation (1). The acquired voltages were then converted to their equivalent FSR As the applied force increases, the FSR resistance also decreases, showing an increased output resistance values by substitution of Equation (1). voltage according to Equation (1). The acquired voltages were then converted to their equivalent FSR 5𝑉 11𝑘𝛺 11𝑘𝛺 𝐹𝑆𝑅 (1). (2) resistance values by substitution of Equation 𝑉 FSR = 4.1.2. Piezo‐Electric Sensor Calibration 5V × 11kΩ − 11kΩ Vout (2) The piezo‐electric sensor generates high voltage values, as high as 20 V with weights less than 5 kg, which requires using a voltage divider circuit before data acquisition by microcontroller. 4.1.2. Piezo-Electric Sensor Calibration V 𝑉𝑜𝑙𝑡𝑎𝑔𝑒 𝑑𝑖𝑣𝑖𝑑𝑒𝑟 𝐺𝑎𝑖𝑛 as Vhigh as 20 V with weights less (3) than 5 The piezo-electric sensor generates high voltage values, kg, which requires using a voltage divider circuit before data acquisition by microcontroller. Vmax input = Voltage divider Gain × VPiezo max ⇒ Voltage divider Gain = Vmax input VPiezo max = VCC VPiezo max = 5V = 0.25 20V (3) (4) Therefore, the voltage divider circuit were chosen as follows: Vout = R1 3 MΩ × VPiezo = × VPiezo = 0.25VPiezo R1 + R2 3 MΩ + 9 MΩ (5) Substituting the maximum piezo voltage in Equation (5) gives: Vout max = 0.25VPiezo max = 0.25(20V ) = 5V (6) This ensures that maximum microcontroller input voltage (5 V) was not exceeded. The acquired voltages were then converted to their equivalent Piezo sensor voltage outputs by subject substitution of Equation (5). 1 VPiezo = × Vout = 4Vout (7) 0.25 Sensors 2020, 20, 957 13 of 25 There are different equivalent electrical models for the piezo-electric sensors [28]. A simplified common model is a voltage source/generator with a capacitance, which was used in this study. Usually, the capacitance values are in Nano Farad range. The equivalent capacitance is typically measured using a parallel connection of a capacitance meter to the sensor. Connecting the piezo-electric sensor to the voltage divider circuit forms a first order high-pass filter. Therefore, high resistance values in mega ohms were used to ensure that most of the generated frequencies by the applied forces would pass. Assuming equivalent capacitance of piezo-electric sensor equal to 9 nF, the cut-off frequency can be written as: 1 1 fcutt−o f f = = = 1.47Hz (8) 2πRC 2π(3M + 9M)9nF Apart from DC and very low frequency components, other signal components were expected to be applied to the MCU input. AD620AN instrumentational amplifiers were used to amplify the low amplitude load cell signals, before it was applied to the microcontroller. The load cells give an output of maximum 40 mV, which can be amplified to the full-scale range of the analog channel. Therefore, the gain of the amplifier and the amplifier gain resistor were chosen as follows: G= Vcc Vload max RG = = 5V = 125 40mV 49.9kΩ 49.9k = = 402 Ω G−1 124 (9) (10) 4.1.3. MEMS Sensor Calibration As mentioned previously, the MEMS generates positive or negative amplitude signals based on the applied force in x, y or z directions. This requires an offset circuit along with a voltage divider circuit to reduce the signal amplitude. It is assumed the piezo-vibration output can go up to 10 V with the maximum applied force. Vmax input = Voltage divider Gain × VPiezo max Voltage divider Gain = Vmax input VPiezo max = VCC /2 5V/2 = = 0.25 VPiezo max 10V (11) (12) Therefore, the voltage divider circuit were chosen as follows: Vout = R1 3 MΩ × VPiezo = × VPiezo = 0.25VPiezo , R1 + R2 3 MΩ + 9 MΩ (13) Substituting the maximum and minimum piezo voltage in Equation (13) gives: Vout max/min = 0.25VPiezo max/min = 0.25(±10V ) = ±2.5V (14) The next step is to add an offset of 1/2Vcc to ensure that the signal was within 0 V to Vcc range. 4.2. Piezo-Electric Sensor Response The piezo-electric sensors can detect the applied forces efficiently, by converting the mechanical movements into electrical signals. However, the movements need to be dynamic. The piezo-electric sensor generated electrical pulses that mimicked the applied mechanical movement. If the mechanical movement was a fast press and release of finger on the active area of the piezo-electric sensor, the pulse was shrunk to an impulse-liked shape. On the other hand, if a gentle force was applied by a slow press and remove by the palm of a hand, the generated signal had irregular pulse shape with longer duration compared to the fast finger press. Even though the piezo-electric sensor’s output can mimic dynamically changing force, it fails to Sensors 2020, 20, 957 14 of 25 The Sensors 2020, 20, 957next step is to add an offset of 1/2Vcc to ensure that the signal was within 0 V to Vcc range. 14 of 25 4.2. Piezo‐Electric Sensor Response detect a static force. Therefore, when applied forceforces is a mixture and force such as The piezo‐electric sensors can the detect the applied efficiently,ofbydynamic converting the static mechanical movements into electrical signals. However, the movements beacquire dynamic.static The piezo‐electric smart insole application, the piezo-electric sensors cannot beneed usedtoto pressure. However, sensor generated electrical that mimicked the applied movement. If theFigure 10 the piezo-electric sensors can bepulses used to detect heel strike or toemechanical off with good accuracy. mechanical movement was a fast press and release of finger on the active area of the piezo‐electric illustrates the individual sensor output for the different applied forces. sensor, the pulse was shrunk to an impulse‐liked shape. Figure 10. Piezo insoleinsole sensors output serial plotter: (A)finger fast finger press and release, Figure 10. Piezo sensors outputwith with Arduino Arduino serial plotter: (A) fast press and release, slowpress palm press and release, sensorsoutput output for walking. (B) slow(B) palm and release, (C)(C) sensors fortwo‐step two-step walking. Onshows the other if a from gentleaforce was applied by a slow press andinsole remove the gait palmcycles. of a Figure 11 thehand, output single piezo-electric sensor of an forbyfew When hand, the generated signal had irregular pulse shape with longer duration compared to the fast finger a force was applied vertically on the sensor’s active area (ceramic), it compressed, exerting an electrical press. Even though the piezo‐electric sensor’s output can mimic dynamically changing force, it fails impulse to with a positive peakTherefore, that mimicked mechanical applied. Theand electrical signal went back detect a static force. when thethe applied force is aforce mixture of dynamic static force such to zero. as Assmart the applied force wasthe released, the signal continued some negative values as the piezo insole application, piezo‐electric sensors cannot be to used to acquire static pressure. piezo‐electric can of be the usedapplied to detectforce. heel strike or toe with good accuracy. ceramic However, bouncedthe to the oppositesensors direction Finally, theoffsignal returned to zero. The Figure 10 illustrates the individual sensor output for the different applied forces. microcontroller clipped the negative part of the signal. However, some part of the negative signal was Figure 11 shows the output from a single piezo‐electric sensor of an insole for few gait cycles. still there, due to the offset added in the acquisition circuit as illustrated in Figure 11. a force SensorsWhen 2020, 20, 957 was applied vertically on the sensor’s active area (ceramic), it compressed, exerting an15 of 25 Voltage Voltage Voltage electrical impulse with a positive peak that mimicked the mechanical force applied. The electrical signal went back to zero. As the applied force was released, the signal continued to some negative 15V values as the piezo ceramic bounced to the opposite direction of the applied force. Finally, the signal returned to zero. The microcontroller clipped the negative part of the signal. 1.3V However, some part of 1V the negative signal was still there, due to the offset added in the acquisition circuit as illustrated in Figure 11. 0.3V 0 t ‐4.8V 0 ‐0.32V Example of a piezo electric sensors output when a vertical force is applied then released ‐0.02V t t Added offset (0.2‐0.4V), this could be added noise picked by the high impedance resistors of the voltage divider (in Mega ohms range) The piezo signal after being attenuated by the voltage divider circuit, assuming a factor of (1/15) 0.3V t 0 The analog input to the microcontroller. The negative part of the signal will be clipped Voltage 1V 1.3V Voltage Voltage 15V 0 t ‐0.30V Removing the offset (‐0.3) 0 t ‐4.5V Multiplying by a gain (x15) to compensate for the attenuation by the voltage divider circuit Voltage The signal will repeat for the next gait cycles, generating a periodic signal. Normally the output during the swing phase is expected to be zero. However, the swing phase intersect with the negative part following the stance phase. That’s why no zero period is shown between the consecutive cycles t Figure11. 11.Mimicking Mimickingpiezoelectric piezoelectric sensor sensor output Figure outputduring duringgait gaitcycle. cycle. 5. Results and Discussion This section illustrates and discuss the results obtained from calibration and characterization tests. Sensors 2020, 20, 957 15 of 25 5. Results and Discussion This section illustrates and discuss the results obtained from calibration and characterization tests. 5.1. Sensors’ Calibration 5.1.1. FSR Sensor Calibration Weight [kg] Weight [kg] Output Voltage [V] Three calibration trials were undertaken for one FSR sensor from Interlink Electronics [22], following the calibration procedure explained previously; 500 g weights where placed one by one every 3–4 s until it reached 5000 g, followed by unloading process from 5000 g down to 0 g. In the loading experiment, output voltage from the voltage divider circuit showed increasing values reflecting the decrease in FSR resistance as shown in Figure 12A. When the applied weight was constant, the output Sensors 2020, 20, 957 constant with small variations. 16 of 25 voltage remained (A) (B) Figure Figure 12. FSR FSR calibration calibration test test (A) (A) applied applied weight weight and and FSR FSR circuit output vs. time time (B) (B) applied applied weight weight FSR resistance. resistance. vs. FSR Weight [kg] In addition, if the constant for a longer time (a few minutes), the sensor voltage For each of the three trials, weight loadingwas andkept off‐loading relationships where obtained. stabilized to atests steady value. However, the aim of this study was to investigate the dynamic response of Loading best fit relations: . the FSR. Therefore, the average output voltage4233.3 for the∗ sample were calculated and plotted against the 𝑊𝑒𝑖𝑔ℎ𝑡 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . corresponding applied weights. Figure 12B 3657.4 shows the plotted data with the fitted waveform. The 𝑊𝑒𝑖𝑔ℎ𝑡 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . calibration showed slight difference the loading and unloading curves, which was expected 𝑊𝑒𝑖𝑔ℎ𝑡between 5117.4 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 due to the hysteresis behavior of FSRs. However, the error was caused by the FSR hysteresis, which can be neglected, as the difference was not significantly high. This can be justified if the response from the smart insole using FSR sensors resembles typical vGRF reported in the literature. Off-Loading tests best fit relations: WeightTrial1 = 5035.2 ∗ Resistance−1.72 WeightTrial2 = 3436.5 ∗ Resistance−1.895 WeightTrial3 = 8111.8 ∗ Resistance−2.589 It is evident that the first and third trial relationships were close to each other (Figure 13). Therefore, either of them can be chosen for the FSR insole. The second trial showed a steaper curve due to higher hystersis error. Figure 13. FSR calibration test: best fit curves between applied weight and FSR resistance for three trials. Off‐Loading tests best fit relations: 𝑊𝑒𝑖𝑔ℎ𝑡 5035.2 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . 𝑊𝑒𝑖𝑔ℎ𝑡 3436.5 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . 𝑊𝑒𝑖𝑔ℎ𝑡 8111.8 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . It is evident that the first and third trial relationships were close to each other (Figure 13). (A) (B) Sensors 2020,Figure 20, 957 12. FSR calibration test (A) applied weight and FSR circuit output vs. time (B) applied weight 16 of 25 vs. FSR resistance. Weight [kg] For each of the three trials, loading and off‐loading relationships where obtained. Loading tests best fit relations: 𝑊𝑒𝑖𝑔ℎ𝑡 4233.3 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . 𝑊𝑒𝑖𝑔ℎ𝑡 3657.4 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . 𝑊𝑒𝑖𝑔ℎ𝑡 5117.4 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . Figure 13.calibration FSR calibration test: fit between curves between appliedand weight and FSR for resistance for three Figure 13. FSR test: best fit best curves applied weight FSR resistance three trials. trials. 5.1.2. Piezo-Electric Sensor Calibration bestwere fit relations: Two Off‐Loading piezoelectrictests sensors used in the calibration process. Three trials were conducted on . 𝑊𝑒𝑖𝑔ℎ𝑡 5035.2 The ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 the first sensors with four trials for the second sensor. piezoelectric sensors showed .a linear . Sensors 2020, 20, 957 𝑊𝑒𝑖𝑔ℎ𝑡 3436.5 ∗ (Figure 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑊𝑒𝑖𝑔ℎ𝑡 8111.8 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 17 of 25 relationship with the applied weights 14). It is evident that the first and third trial relationships were close to each other (Figure 13). Therefore, either of them can be chosen for the FSR insole. The second trial showed a steaper curve Reference Mass vs Time due to higher hystersis error. 6 4 5.1.2. Piezo‐Electric Sensor Calibration 2 Weight [kg] Two piezoelectric sensors were used in the calibration process. Three trials were conducted on the sensor. The piezoelectric sensors showed a linear 0 first 2 sensors 4 6 with 8 four 10 trials 12 for 14 the 16 second 18 Time [sec] relationship with the applied weights (Figure 14). 0 Piezo Voltage vs Time 15 10 5 0 0 2 4 6 8 10 Time [sec] 12 14 16 18 (A) (B) Figure 14. 14. Piezo‐electric calibration test: test: (A) weight vs. vs. time output Figure Piezo-electric sensor sensor calibration (A) applied applied weight time and and piezoelectric piezoelectric output voltage vs. time (B) applied weight vs. piezoelectric output voltage. voltage vs. time (B) applied weight vs. piezoelectric output voltage. trials for for the the 2nd 2nd piezo The second and fourth trials piezo sensor sensor had had different different slopes compared to the trials. This This could couldbe berelated relatedtotothe thecalibration calibrationprocess process itself, weights were applied remaining trials. itself, asas thethe weights were applied by by fast presses and releases active area thesensor. sensor.Therefore, Therefore,applying applyingthe theforce force on on the the exact fast presses and releases onon thethe active area ofofthe guaranteed between successive readings. This same areas is not guaranteed This issue issue can can be overcome by using a machine to apply the weights. However, this would defeat the purpose of the study in providing a low‐cost setup (Figure 15). low-cost 6 5 4 Weight vs Piezo-electric Voltage remaining trials. This could be related to the calibration process itself, as the weights were applied by fast presses and releases on the active area of the sensor. Therefore, applying the force on the exact same areas is not guaranteed between successive readings. This issue can be overcome by using a machine to apply the weights. However, this would defeat the purpose of the study in providing a Sensors 2020, 20, 957 17 of 25 low‐cost setup (Figure 15). Weight vs Piezo-electric Voltage 6 5 4 3 Piezo1 Trial1 Piezo1 Trial2 Piezo1 Trial3 Piezo2 Trial1 Piezo2 Trial2 Piezo2 Trial3 Piezo2 Trial4 2 1 0 -1 0 2 4 6 8 Voltage [V] 10 12 14 Figure 15. 15. Piezoelectric test:seven seventrials trials shows relationship between the applied Figure Piezoelectricsensor sensorcalibration calibration test: shows relationship between the applied weight piezoelectricoutput output voltage. voltage. weight andand piezoelectric Obtained lines of best fit: Obtained lines of best fit: 𝑊𝑒𝑖𝑔ℎ𝑡 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.19123 WeightPiezo1Trial1 = 0.42867 0.42867 ∗ ∗PiezoVoltage − 0.19123 𝑊𝑒𝑖𝑔ℎ𝑡 0.41110 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.0081012 𝑊𝑒𝑖𝑔ℎ𝑡 0.39321∗ PiezoVoltage ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.084656 WeightPiezo1Trial2 = 0.41110 + 0.0081012 𝑊𝑒𝑖𝑔ℎ𝑡 0.34619 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.4105 WeightPiezo1Trial3 = 0.39321 ∗ PiezoVoltage + 0.084656 𝑊𝑒𝑖𝑔ℎ𝑡 0.27242 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.57351 WeightPiezo2Trial1 =0.35564 0.34619 ∗∗PiezoVoltage + 0.4105 𝑊𝑒𝑖𝑔ℎ𝑡 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.30325 𝑊𝑒𝑖𝑔ℎ𝑡 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.36416 WeightPiezo2Trial2 = 0.31765 0.27242 ∗ ∗ PiezoVoltage + 0.57351 WeightPiezo2Trial3 = 0.35564 ∗ PiezoVoltage + 0.30325 5.1.3. MEMS Sensor Calibration WeightPiezo2Trial4 = 0.31765 ∗ PiezoVoltage + 0.36416 5.1.3. MEMS Sensor Calibration Twenty different calibration trials were conducted on a piezo-vibration sensor. However, high repeatability error persisted, making it difficult to obtain a clear relation between sensor output voltage and the applied weight. The applied weight showed a direct proportional relation with output voltage for some successive readings and an inverse relation with some other successive readings. This is because of the MEMS sensitivity to the applied force in 3-D space (x, y or z directions). It generates 1-D output voltage with positive or negative amplitude depending on the applied force in certain direction. Therefore, if the applied force is a summation of forces in 2 or 3 axes, the output voltage might go to zero or attenuated with the addition of different sign amplitudes. A linear relation was not clearly obtained by the application of vertical forces, as the applied force might not be applied in one axis only (Figure 16). readings. This is because of the MEMS sensitivity to the applied force in 3‐D space (x, y or z directions). It generates 1‐D output voltage with positive or negative amplitude depending on the applied force in certain direction. Therefore, if the applied force is a summation of forces in 2 or 3 axes, the output voltage might go to zero or attenuated with the addition of different sign amplitudes. A2020, linear relation was not clearly obtained by the application of vertical forces, as the applied Sensors 20, 957 18 of 25 force might not be applied in one axis only (Figure 16). Sensors 2020, 20, 957 Reference Mass vs Time 18 of 25 4 Amplitude [V] Weight [kg] Twenty different calibration trials were conducted on a piezo‐vibration sensor. However, high repeatability error persisted, making it difficult to obtain a clear relation between sensor output 2 voltage and the applied weight. The applied weight showed a direct proportional relation with output voltage for some successive readings and an inverse relation with some other successive 0 0 5 10 20 25 readings. This is because of15the MEMS sensitivity to the applied force in 3‐D space (x, y or z Time [sec] directions). It generates 1‐D output voltage with positive or negative amplitude depending on the Piezo-vibration Voltage vs Time applied force in certain direction. Therefore, if the applied force is a summation of forces in 2 or 3 4 axes, the output voltage might go to zero or attenuated with the addition of different sign amplitudes. A linear relation was not clearly obtained by the application of vertical forces, as the applied 2 force might not be applied in one axis only (Figure 16). 0 Weight [kg] 0 5 10 15 Reference Mass vs Time Time [sec] 4 20 25 (a) (b) 2 Figure 16. 16. MEMS MEMSsensor sensor calibration calibration test: test: (A) (A)applied applied weight weight vs. vs. time timeand andMEMS MEMSsensor sensoroutput output voltage voltage vs. time time (B) (B) applied applied weight weight vs. vs. MEMS MEMS sensor sensor output output voltage. voltage. 0 0 5 0 5 10 15 20 25 20 25 0 Weight [kg] Amplitude [V] The mathematical relations Time [sec] obtained in the calibration phase cannot be used to design a piezo-vibration sensor-based smart insole, since the applied force in gait can be in any of the x, Piezo-vibration Voltage vs Time y or z4directions (Figure 17). Therefore, the piezo-vibration sensor was discarded from the sensor list for designing smart insole. However, it can be utilized in other biomedical applications where the force 2directions are limited to a certain axis or a fixed plane. Moreover, it can be used to detect initial timing of the applied force. The lines of best fit obtained were: 10 15 TimeWeight [sec] Trial1 = 1.4145 ∗ PiezoVoltage − 1.3447 (a) WeightTrial2 = 1.5215 ∗ PiezoVoltage − 1.8581 (b) Figure 16. MEMS sensor calibration test: (A) applied weight vs. time and MEMS sensor output voltage WeightTrial3 = 0.80343 ∗ PiezoVoltage − 0.22721 vs. time (B) applied weight vs. MEMS sensor output voltage. Figure 17. MEMS sensor calibration test: 3 best trials shows relationship between the applied weight and MEMS sensor’s output voltage. Weight [kg] The mathematical relations obtained in the calibration phase cannot be used to design a piezo‐ vibration sensor‐based smart insole, since the applied force in gait can be in any of the x, y or z directions (Figure 17). Therefore, the piezo‐vibration sensor was discarded from the sensor list for designing smart insole. However, it can be utilized in other biomedical applications where the force directions are limited to a certain axis or a fixed plane. Moreover, it can be used to detect initial timing of the applied force. The lines of best fit obtained were: Figure MEMSsensor sensorcalibration calibrationtest: test:3 3best besttrials trialsshows showsrelationship relationshipbetween betweenthe the applied weight Figure 17.17.MEMS applied weight and MEMS sensor’soutput outputvoltage. voltage. and MEMS sensor’s The mathematical relations obtained in the calibration phase cannot be used to design a piezo‐ vibration sensor‐based smart insole, since the applied force in gait can be in any of the x, y or z directions (Figure 17). Therefore, the piezo‐vibration sensor was discarded from the sensor list for designing smart insole. However, it can be utilized in other biomedical applications where the force directions are limited to a certain axis or a fixed plane. Moreover, it can be used to detect initial timing of the applied force. The lines of best fit obtained were: Sensors 2020, 20, 957 Sensors 2020, 20, 957 19 of 25 𝑊𝑒𝑖𝑔ℎ𝑡 𝑊𝑒𝑖𝑔ℎ𝑡 𝑊𝑒𝑖𝑔ℎ𝑡 1.4145 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 1.5215 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.80343 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 1.3447 1.8581 0.22721 19 of 25 5.2. Insole Charecterization 5.2. Insole Charecterization 5.2.1. FSR-Based Insole Characterization 5.2.1. FSR‐Based Insole Characterization The gait cycle ofof 1212 subjects were recorded while walking onon a 10 mm walkway in in self-selected The gait cycle subjects were recorded while walking a 10 walkway self‐selected walking manner. Each subject had 6 trials recorded, where the first and last few (1(1 toto 3) 3) cycles were walking manner. Each subject had 6 trials recorded, where the first and last few cycles were discarded from cycles for forboth bothfeet feetwere wereconsidered consideredfor discarded fromeach eachtrial, trial,and andthe theremaining remaining part part of the gait cycles foranalysis. analysis. The cycle of one of the subjects is analyzed in following the following illustrating a The gaitgait cycle of one of the subjects is analyzed in the lines,lines, illustrating a simple simple analysis technique replicated different application application by researchers analysis technique that that can can be be replicated inindifferent researchersworking workingwith with wearable insoles. wearable insoles. InIn normal gait cycles both heel peak (first peak) and toe offoff (peak) must show close values, with normal gait cycles both heel peak (first peak) and toe (peak) must show close values, with both feet having symmetrical signals. signal showed close peak values both feet having symmetrical signals.Even Eventhough thoughthe theright rightfoot foot signal showed close peak values (Figure 18), thethe left foot signals showed a big variance between thethe heel-strike and toe-off peaks. This (Figure 18), left foot signals showed a big variance between heel‐strike and toe‐off peaks. This can bebe explained, the sensitivity between the FSRs ofof the insole and their hysteresis can explained,byby the sensitivitydifference difference between the FSRs the insole and their hysteresis effect. AsAs explained earlier, this issue was mitigated by some teams using a regression models effect. explained earlier, this issue was mitigated byresearch some research teams using a regression that calibrates the FSR insole readings a reference signal, recorded in motionin models that calibrates the FSR insoleagainst readings against a reference signal, simultaneously recorded simultaneously analysis [34,35]. This expensive be neglected some applications, where the quality motionlabs analysis labs [35,50]. This approach expensivecan approach can beinneglected in some applications, where ofthe thequality acquired is sufficient toisachieve theto desired goal. instance, detection application, ofsignal the acquired signal sufficient achieve the For desired goal.smart For instance, smart detection where the machine-learning algorithm can differentiate different groups people even withof application, where the machine‐learning algorithm between can differentiate betweenofdifferent groups lowor medium-quality gait cycles recorded (vGRF). gait cycles (vGRF). people even with low‐ recorded or medium‐quality Gait Cycles 50 Left Foot Right Foot 40 30 20 10 0 0 1 2 Time [sec] 3 4 Figure Gait cycles readings left and right foot with FSR smart insole. Figure 18.18. Gait cycles readings forfor left and right foot with FSR smart insole. The full gait record was segmented into distinct gait cycles. Then was resampled into 512 The full gait record was segmented into distinct gait cycles. Then it it was resampled into toto 512 sample.Segmentation Segmentationisisa acommon commonpractice practicetotofacilitate facilitatethe thecomparison comparisonbetween betweenallallthe thegait gaitcycles. cycles. sample. The segmented gait cycles used smart detection algorithms where segments equal length are The segmented gait cycles areare used inin smart detection algorithms where segments ofof equal length are used train specific machine learning algorithms to classify different groups of people based used to to train specific machine learning algorithms to classify different groups of people based on theiron their In addition, statistical the segmented such as mean, standard deviation, gait. Ingait. addition, statistical data ofdata the of segmented cyclescycles such as mean, standard deviation, timetime to to peaks, and percentage of stance phase in a full cycle/stride (stance phase plus swing phase) can peaks, and percentage of stance phase in a full cycle/stride (stance phase plus swing phase) can bebe utilized a gait analysis tool sports and medical applications. utilized asas a gait analysis tool in in sports and medical applications. Thesegmentation segmentationwas wascarried carriedout outbybya acustomized customizedMATLAB MATLABcode codethat thatdetects detectsgroups groupsofof The consecutivenon-zero non‐zero samples.Then Thenit itsegments segmentsthose thosesignals signalsinto into individualstance stance phases, each consecutive samples. individual phases, each starting with a heel‐strike and ending with a toe off. Figure shows the first 10‐m trial one the starting with a heel-strike and ending with a toe off. Figure 1818 shows the first 10-m trial ofof one ofof the participants, where four gait cycles were extracted after excluding the first and last two gait cycles. Left foot vGRF was segmented into four stance phases (Figure 19A). Sensors 2020, 957 Sensors 2020, 20,20, 957 2020 of of 2525 Sensors 2020, 20, 957 20 of 25 participants,where wherefour fourgait gaitcycles cycleswere wereextracted extractedafter afterexcluding excludingthe thefirst firstand andlast lasttwo twogait gaitcycles. cycles. participants, Left foot vGRF was segmented into four stance phases (Figure 19A). Left foot vGRF was segmented into four stance phases (Figure 19A). (A) (A) (B) (B) Figure (A) Segmented left‐foot gait cycles, (B) segmented right‐foot cycles. Figure 19. (A) Segmented left‐foot gait cycles, (B) segmented right‐foot cycles. Figure 19.19. (A) Segmented left-foot gait cycles, (B) segmented right-foot cycles. vGRF [kg] vGRF [kg] The data were sampled with asampling sampling rate samples/second, where each segment (stance The data were sampled with rate ofof 6060 samples/second, where each segment (stance The data were sampled with aa sampling rate of 60 samples/second, where each segment (stance phase)takes takes around Therefore,each eachsegment segmentconsists consistsof around42 samples,which which were then phase) takes around Therefore, each segment consists ofof around 4242 samples, which were then phase) around 0.70.7s. s. Therefore, around samples, were then resampled into 512 samples. The mean values and standard deviations of each of the 512 samples resampled into and standard deviations of each of the with resampled into 512 512samples. samples.The Themean meanvalues values and standard deviations of each of 512 the samples 512 samples withrespect respect the4 4segmented segmented signals werecalculated. calculated. Thenthe themean mean values along withthe the respect to theto 4to segmented signals signals were calculated. Then theThen mean values along with the deviation with the were values along with deviation from the means (means plus and minus the deviation) the left foot was calculated and from the means (means plus and plus minus the deviation) for the left foot was calculated and plotted deviation from the means (means and minus the deviation) forfor the left foot was calculated and plotted20). (Figure 20). Similar steps wererepeated repeated with leftfoot footvGRF vGRF (Figure 20).This This providesanan (Figure Similar steps were repeated with left foot vGRF (Figure 20).(Figure This provides anprovides illustrative plotted (Figure 20). Similar steps were with left 20). illustrative figure thatcan canbe usedbybyinsport indifferent different sportand and medicalapplications applications asses walking figure that figure can bethat used by inbeused different and medical applications to asses walking behaviors illustrative sport medical totoasses walking complications. orbehaviors complications. behaviors oror complications. Figure Means and standard deviations of gait cycles; blue curves represents the mean gait value Figure 20.20. Means and deviations ofgait gait cycles; blue curves represents the mean gait value Figure Means standard deviations of cycles; blue curves represents the mean gait value of ofthe theleft leftfoot footwith withdashed dashed linerepresenting representing thedeviation deviation fromthe the mean, while orange curves of line the mean, while orange curves the left foot with dashed line representing the deviation from thefrom mean, while orange curves represents represents the mean gait value of the right foot with dashed line representing the deviation from the represents the mean gait value of the right foot with dashed line representing deviation from the the mean gait value of the right foot with dashed line representing the deviationthe from the mean value. mean value. mean value. The vGRF of a subject mainly depends on his/her health condition and the footwear used. In this study, all participants were advised to wear comfortable walking shoes avoiding high-heel shoes, The vGRF a subject mainly depends his/her health condition and the footwear used. In this The vGRF ofof a subject mainly depends onon his/her health condition and the footwear used. In this especially for female subjects. This ensured that all subjects went shoes through similar high‐heel condition while study,allallparticipants participants wereadvised advised wear comfortable walking shoesavoiding avoiding high‐heelshoes, shoes, study, were totowear comfortable walking conducting the experiment. ItThis was observed that the collected data did not showcondition any significant especiallyforfor female subjects. Thisensured ensuredthat that subjects went through similar condition while especially female subjects. allall subjects went through similar while statistical difference based on gender. Sensors 2020, 20, 957 21 of 25 Sensors 2020, 20, 957 21 of 25 conducting the experiment. It was observed that the collected data did not show any significant statistical difference based on gender. 5.2.2. Piezoelectric Sensor Based-Insole Characterization 5.2.2. Piezoelectric Sensor Based‐Insole Characterization Three subjects participated in the piezo-electric insole test in the same manner as the testing of Three subjects participated in the piezo‐electric insole test in the same manner as the testing of FSR based insole. The piezoelectric insoles were expected to detect the gait cycle, with impulse signals FSR based insole. The piezoelectric insoles were expected to detect the gait cycle, with impulse signals in heel-area sensors during the heel strike phase and lower amplitude impulses from all the sensors in heel‐area sensors during the heel strike phase and lower amplitude impulses from all the sensors during the mid-stance phase. Finally, impulse signals from the toe and metatarsal heads sensors were during the mid‐stance phase. Finally, impulse signals from the toe and metatarsal heads sensors were taken in the toe-off phase. However, the readings were not promising, showing single irregular shape taken in the toe‐off phase. However, the readings were not promising, showing single irregular shape pulses per sensor for each individual gait cycle. The addition of different sensors output showed pulses per sensor for each individual gait cycle. The addition of different sensors output showed periodic impulses, one impulse per period (Figure 21). This indicates that the full stance period was periodic impulses, one impulse per period (Figure 21). This indicates that the full stance period was detected as one event only. Meanwhile, the correct vGRF must show two distinct peaks between detected as one event only. Meanwhile, the correct vGRF must show two distinct peaks between the the mid-stance phase, summing up to three main phases: heel strike, mid stance and toe off. The mid‐stance phase, summing up to three main phases: heel strike, mid stance and toe off. The rigid rigid nature of the piezo sensor made it difficult to detect different gait phases. Therefore, it can be nature of the piezo sensor made it difficult to detect different gait phases. Therefore, it can be summarized that it is not suitable for the smart insole application which requires to produce reliable summarized that it is not suitable for the smart insole application which requires to produce reliable vGRF signal due to gait. vGRF signal due to gait. Gait Cycles (Subject A) 100 Left Foot Right Foot vGRF [kg] 50 0 -50 0 1 2 3 4 Time [sec] (a) 5 6 7 (b) Figure21. 21.Gait Gaitcycles cyclesfor forleft leftand andright rightfoot footwith withpiezoelectric piezoelectricsmart smartinsole insole(a)(A) subject 1, (B) subject Figure subject 1, (b) subject 2. 2. In this study, the authors have characterized three samples from each sensor category randomly; however, thestudy, smartthe insole was have implemented usingthree 16-sensors. was expected there In this authors characterized samplesTherefore, from each it sensor category that randomly; would be a the small variation the implemented vGRF recordedusing from 16‐sensors. the smart insole in different trials and inthat different however, smart insoleofwas Therefore, it was expected there subjects. However, 20 clearly that the vGRF individual has a unique would be a small Figure variation of thedepicts vGRF recorded fromfrom the an smart insole foot in different trialspattern and in and this finding matches with the vGRF by the commercial smart insole and force plate. Thisa different subjects. However, Figure 20recorded clearly depicts that the vGRF from an individual foot has reflects fact that insole designed FSR is capable by of acquiring vGRF smart reliably and the uniquethe pattern andthe thissmart finding matches withusing the vGRF recorded the commercial insole and designed system robust enough adapt the age, gender and body mass (BMI) of variation of force plate. Thisisreflects the fact to that the to smart insole designed using FSRindex is capable acquiring the participants. However, carbon piezoresistive material (like Velostat), which authors have tested in vGRF reliably and the designed system is robust enough to adapt to the age, gender and body mass preliminary here However, in order to carbon avoid unnecessary length of the(like manuscript), index (BMI)experiments variation of(not the reported participants. piezoresistive material Velostat), showed highhave hysteresis type of material is not suitable for humanhere dynamicity monitoring. which very authors testedand inthis preliminary experiments (not reported in order to avoid On the other hand, sensors showed can monitor pressure variation theyisare unnecessary lengthpiezoelectric of the manuscript), very dynamic high hysteresis and this typehowever, of material not very sensitive to smalldynamicity pressure change and incapable reliably produce mean vGRF. Moreover, the suitable for human monitoring. On the to other hand, piezoelectric sensors can monitor vGRF changed overvariation trials and over subjects significantly and, temporal feature of vGRF dynamic pressure however, they are very sensitive totherefore, small pressure change and incapable cannot be identified piezoelectric sensor-based to reliably produceusing mean vGRF. Moreover, the smart vGRF insole. changed over trials and over subjects significantly and, therefore, temporal feature of vGRF cannot be identified using piezoelectric sensor‐ 5.3. Performance Evaluation of the FSR-Based System based smart insole. Comparing the mean and standard deviation of vGRF for a gait cycle of the same subject recorded 5.3. Performance Evaluation of the FSR‐Based using the two systems, commercial F-scan System and the proposed FSR insoles, it can be seen that both showed good quality signals except slight differencesofinvGRF peak for values (Figure Thesame FSR subject insole Comparing the mean and standard deviation a gait cycle 22). of the showed smaller vGRF during left-foot heel strike phase compared to the F-scan insole. This isthat an recorded using the two systems, commercial F‐scan and the proposed FSR insoles, it can be seen expected behavior as each sensor is somewhat unique due to the manufacturing process and we cannot Sensors 2020, 20, 957 22 of 25 both showed good quality signals except slight differences in peak values (Figure 22). The FSR insole showed smaller vGRF during left‐foot heel strike phase compared to the F‐scan insole. This Sensors 2020, 20, 957 22 ofis25an expected behavior as each sensor is somewhat unique due to the manufacturing process and we cannot calibrate individual sensors, which can lead to some variation. In addition, due to the presence calibrate individual sensors, of which lead to somemore variation. In addition, dueshoe. to the of an insole, the sensitivity somecan FSRs decreases than the others in the Topresence mitigate of this anproblem, insole, the sensitivity of some FSRs decreases more than the others in the shoe. To mitigate this a highly uniform pressure should be applied across individual sensors. Each sensor should problem, uniform pressure across sensor should producea highly uniform output. When should this is be notapplied the case for individual a specific sensors. sensor, Each the software should produce uniform output. When this is not the case for a specific sensor, the software should determine determine a unique scale factor to compensate for the output variation. Currently, there are a few a unique scalesuch factorastoT‐scan, compensate for the output variation. Currently, there are a few companies companies that provide a special piece of equipment (equilibration device) such which asapplies T-scan, athat provide a special piece of equipment (equilibration device) which applies a uniform uniform pressure on the full insole using a thin flexible membrane to perform such pressure on the full insole using a to thin perform such calibration. calibration. Moreover, compared theflexible F‐scan membrane system, theto FSR readings showed smallerMoreover, differences compared the F-scan system, FSRvalues. readings smaller between vGRF between to vGRF peaks and mid the stance Thisshowed was mainly duedifferences to the superior number ofpeaks sensors and mid stance values. This was mainly due to the superior number of sensors for the F-scan system for the F‐scan system (960 sensing areas) compared to the proposed insole (16 FSRs). In addition, the (960 sensing areas)elements compared to the proposed insole (16 on FSRs). In addition, the F-scan elements F‐scan sensing were uniformly distributed the full foot area, while thesensing FSR sensors were were uniformly on the most full foot area, while the FSR sensors placedplaced on the foot areas placed on the distributed foot areas where of the pressure is exerted withwere no sensors on the low‐ where most of the pressure is exerted with no sensors placed on the low-pressure areas (medial arch). pressure areas (medial arch). Adding a few FSR sensors to the medial arch can improve the quality Adding few FSR sensors to the medial can improve thefoot, quality of the signal obtained, especially of theasignal obtained, especially forarch subjects with flat who exerts considerable amount of forpressure subjectson with flat foot, who exerts considerable amount of pressure on medial arch areas. medial arch areas. A B Figure Comparison between mean and standard deviation vertical ground reaction forces Figure 22.22. Comparison between thethe mean and standard deviation of of vertical ground reaction forces (vGRF) from (blue) and right (orange) foot using F‐scan system and FSR‐system (vGRF) from leftleft (blue) and right (orange) foot using F-scan system (A)(A) and FSR-system (B).(B). 6. 6. Conclusions Conclusion InIn this study, thethe authors have proposed and designed low-cost calibration setups forfor calibrating this study, authors have proposed and designed low‐cost calibration setups calibrating three different force sensors: FSR, ceramic piezoelectric and flexible piezoelectric sensors. three different force sensors: FSR, ceramic piezoelectric and flexible piezoelectric sensors.The The experiments experimentsconducted conductedshowed showedthe theeffectiveness effectivenessofofthe theproposed proposedsetup setupinincalibrating calibratingFSR FSRand and piezoelectric sensors, which are mainly affectedaffected by 1D force. It was found flexible piezoelectric piezoelectric sensors, which are mainly by 1D force. It that was the found that the flexible sensors were performing poor in terms of calibration due to their sensitivity to 3D forces. Special piezoelectric sensors were performing poor in terms of calibration due to their sensitivity to 3D forces. force calibration machines are required to control the applied force in x, y or z directions. In addition, Special force calibration machines are required to control the applied force in x, y or z directions. In a systematic designing characterizing two differenttwo smart insolessmart were insoles illustrated. addition, aprocedure systematicfor procedure forand designing and characterizing different were The vGRF signal acquired and segmented to obtain mean vGRF and its standard deviation for a gait illustrated. The vGRF signal acquired and segmented to obtain mean vGRF and its standard deviation cycle can be used measure different statistical metricsstatistical (such as mean standard for were a gaitcalculated, cycle werewhich calculated, whichtocan be used to measure different metrics (such as deviation, time to peak, etc.) that can help in assessing the walking behavior of athletes, patients orof mean standard deviation, time to peak, etc.) that can help in assessing the walking behavior normal people. The FSR-based smart insole was able to acquire high quality vGRF for different gait athletes, patients or normal people. The FSR‐based smart insole was able to acquire high quality cycles. other hand, the piezoelectric sensor-based insole failedsensor‐based to detect distinct phases. It vGRFOn forthe different gait cycles. On the other hand, the piezoelectric insolegait failed to detect cannot be utilized as an alternative to FSR in smart insole application. However, the calibrated piezo distinct gait phases. It cannot be utilized as an alternative to FSR in smart insole application. sensors can be in other bio-sensing technologies as detecting the start and end ofsuch eachas However, theutilized calibrated piezo sensors can be utilizedsuch in other bio‐sensing technologies gait cycle. the start and end of each gait cycle. detecting Sensors 2020, 20, 957 23 of 25 Author Contributions: Experiments were designed by M.E.H.C., N.E. and A.K. Experiments were performed and experimental data were acquired by A.M.T., S.A.-H., M.A., and S.A. Data were analyzed by A.M.T., M.E.H.C., A.K. and N.A.-E. All authors were involved in interpretation of data and writing the paper. All authors have read and agreed to the published version of the manuscript. Funding: This research was partially funded by Qatar National Research Foundation (QNRF), grant number NPRP12S-0227-190164 and Research University Grant DIP-2018-017. The publication of this article was funded by the Qatar National Library. Acknowledgments: The authors would like to thank Engr. Ayman Ammar, Electrical Engineering, Qatar University for helping in printing the printed circuit boards (PCBs). Conflicts of Interest: The authors declare no conflict of interest. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. Kowalski, E.; Catelli, D.S.; Lamontagne, M. Side does not matter in healthy young and older individuals–Examining the importance of how we match limbs during gait studies. Gait Posture 2019, 67, 133–136. Tao, W.; Liu, T.; Zheng, R.; Feng, H. Gait analysis using wearable sensors. Sensors 2012, 12, 2255–2283. [CrossRef] [PubMed] Howell, A.M.; Kobayashi, T.; Hayes, H.A.; Foreman, K.B.; Bamberg, S.J.M. Kinetic gait analysis using a low-cost insole. IEEE Trans. Biomed. Eng. 2013, 60, 3284–3290. [CrossRef] [PubMed] Bamberg, S.J.M.; Benbasat, A.Y.; Scarborough, D.M.; Krebs, D.E.; Paradiso, J.A. Gait analysis using a shoe-integrated wireless sensor system. IEEE Trans. Inf. Technol. Biomed. 2008, 12, 413–423. [CrossRef] [PubMed] De Rossi, S.; Lenzi, T.; Vitiello, N.; Donati, M.; Persichetti, A.; Giovacchini, F.; Vecchi, F.; Carrozza, M.C. Development of an in-shoe pressure-sensitive device for gait analysis. In Proceedings of the 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Boston, MA, USA, 30 August–3 September 2011; IEEE: Piscataway, NJ, USA, 2011. Smart Tracking and Wearables: Techniques in Gait Analysis and Movement in Pathological Aging. Available online: https://www.intechopen.com/books/smart-healthcare/smart-tracking-and-wearables-techniquesin-gait-analysis-and-movement-in-pathological-aging (accessed on 9 February 2020). Gurchiek, R.D.; Choquette, R.H.; Beynnon, B.D.; Slauterbeck, J.R.; Tourville, T.W.; Toth, M.J.; McGinnis, R.S. Remote gait analysis using wearable sensors detects asymmetric gait patterns in patients recovering from ACL reconstruction. In Proceedings of the 2019 IEEE 16th International Conference on Wearable and Implantable Body Sensor Networks (BSN), Chicago, IL, USA, 19–22 May 2019; IEEE: Piscataway, NJ, USA, 2019. Brognara, L.; Palumbo, P.; Grimm, B.; Palmerini, L. Assessing gait in parkinson’s disease using wearable motion sensors: A systematic review. Diseases 2019, 7, 18. [CrossRef] [PubMed] Mills, P.M.; Morrison, S.; Lloyd, D.G.; Barrett, R.S. Repeatability of 3D gait kinematics obtained from an electromagnetic tracking system during treadmill locomotion. J. Biomech. 2007, 40, 1504–1511. [CrossRef] Huang, E.; Sharp, M.T.; Osborn, E.; MacLellan, A.; Mlynash, M.; Kemp, S.; Buckwalter, M.S.; Lansberg, M.G. Abstract TP173: Feasibility and Utility of Home-Based Gait Analysis Using Body-Worn Sensors. Stroke 2019, 50, ATP173. [CrossRef] Liu, T.; Inoue, Y.; Shibata, K. A wearable ground reaction force sensor system and its application to the measurement of extrinsic gait variability. Sensors 2010, 10, 10240–10255. [CrossRef] [PubMed] Razak, A.; Hadi, A.; Zayegh, A.; Begg, R.K.; Wahab, Y. Foot plantar pressure measurement system: A review. Sensors 2012, 12, 9884–9912. [CrossRef] [PubMed] Alam, U.; Riley, D.R.; Jugdey, R.S.; Azmi, S.; Rajbhandari, S.; D’Août, K.; Malik, R.A. Diabetic neuropathy and gait: A review. Diabetes Ther. 2017, 8, 1253–1264. [CrossRef] [PubMed] Farahpour, N.; Jafarnezhad, A.; Damavandi, M.; Bakhtiari, A.; Allard, P. Gait ground reaction force characteristics of low back pain patients with pronated foot and able-bodied individuals with and without foot pronation. J. Biomech. 2016, 49, 1705–1710. [CrossRef] [PubMed] Sacco, I.C.; Picon, A.P.; Macedo, D.O.; Butugan, M.K.; Watari, R.; Sartor, C.D. Alterations in the lower limb joint moments precede the peripheral neuropathy diagnosis in diabetes patients. Diabetes Technol. Ther. 2015, 17, 405–412. Sensors 2020, 20, 957 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 24 of 25 Dabbah, M.A.; Graham, J.; Petropoulos, I.N.; Tavakoli, M.; Malik, R.A. Automatic analysis of diabetic peripheral neuropathy using multi-scale quantitative morphology of nerve fibres in corneal confocal microscopy imaging. Med. Image Anal. 2011, 15, 738–747. [CrossRef] [PubMed] International Diabetes Federation: IDF Diabetes Atlas. Available online: https://www.idf.org/e-library/ epidemiology-research/diabetes-atlas/19-atlas-6th-edition.html (accessed on 9 February 2020). Dworkin, R.H.; Malone, D.C.; Panarites, C.J.; Armstrong, E.P.; Pham, S.V. Impact of postherpetic neuralgia and painful diabetic peripheral neuropathy on health care costs. J. Pain 2010, 11, 360–368. [CrossRef] Dyck, P.J.; Overland, C.J.; Low, P.A.; Litchy, W.J.; Davies, J.L.; Dyck, P.J.B.; O’brien, P.C. Signs and symptoms versus nerve conduction studies to diagnose diabetic sensorimotor polyneuropathy: Cl vs. NPhys trial. Muscle Nerve 2010, 42, 157–164. Taksande, B.; Ansari, S.; Jaikishan, A.; Karwasara, V. The diagnostic sensitivity, specificity and reproducibility of the clinical physical examination signs in patients of diabetes mellitus for making diagnosis of peripheral neuropathy. J. Endocrinol. Metab. 2011, 1, 21–26. [CrossRef] Watari, R.; Sartor, C.D.; Picon, A.P.; Butugan, M.K.; Amorim, C.F.; Ortega, N.R.; Sacco, I.C. Effect of diabetic neuropathy severity classified by a fuzzy model in muscle dynamics during gait. J. Neuroeng. Rehabil. 2014, 11, 11. [CrossRef] England, J.; Gronseth, G.; Franklin, G.; Carter, G.; Kinsella, L.; Cohen, J.; Asbury, A.; Szigeti, K.; Lupski, J.; Latov, N. Practice Parameter: Evaluation of distal symmetric polyneuropathy: Role of laboratory and genetic testing (an evidence-based review) Report of the American Academy of Neurology, American Association of Neuromuscular and Electrodiagnostic Medicine, and American Academy of Physical Medicine and Rehabilitation. Neurology 2009, 72, 185–192. Zhang, Q.; Wang, Y.L.; Xia, Y.; Wu, X.; Kirk, T.V.; Chen, X.D. A low-cost and highly integrated sensing insole for plantar pressure measurement. Sens. Bio-Sens. Res. 2019, 26, 100298. [CrossRef] Zhang, W.; Yin, B.; Wang, J.; Mohamed, A.; Jia, H. Ultrasensitive and wearable strain sensors based on natural rubber/graphene foam. J. Alloys Compd. 2019, 785, 1001–1008. [CrossRef] Sengupta, D.; Pei, Y.; Kottapalli, A.G.P. Ultralightweight and 3D Squeezable Graphene-Polydimethylsiloxane Composite Foams as Piezoresistive Sensors. ACS Appl. Mater. Interfaces 2019, 11, 35201–35211. [CrossRef] [PubMed] Shu, L.; Hua, T.; Wang, Y.; Li, Q.; Feng, D.D.; Tao, X. In-shoe plantar pressure measurement and analysis system based on fabric pressure sensing array. IEEE Trans. Inf. Technol. Biomed. 2010, 14, 767–775. [PubMed] Giovanelli, D.; Farella, E. Force sensing resistor and evaluation of technology for wearable body pressure sensing. J. Sens. 2016, 2016. [CrossRef] Feasibility of Force Detection in 3D Printed Flexible Material Using Embedded Sensors. Available online: https://www.spiedigitallibrary.org/conference-proceedings-of-spie/10970/109702F/Feasibilityof-force-detection-in-3D-printed-flexible-material-using/10.1117/12.2514051.short?SSO=1 (accessed on 9 February 2020). Heng, W.; Pang, G.; Xu, F.; Huang, X.; Pang, Z.; Yang, G. Flexible Insole Sensors with Stably Connected Electrodes for Gait Phase Detection. Sensors 2019, 19, 5197. [CrossRef] Lee, W.; Hong, S.-H.; Oh, H.-W. Characterization of Elastic Polymer-Based Smart Insole and a Simple Foot Plantar Pressure Visualization Method Using 16 Electrodes. Sensors 2019, 19, 44. [CrossRef] Theodosiou, A.; Kalli, K. Recent trends and advances of fibre Bragg grating sensors in CYTOP polymer optical fibres. Opt. Fiber Technol. 2020, 54, 102079. [CrossRef] Abdelhady, M.; van den Bogert, A.; Simon, D. A High-Fidelity Wearable System for Measuring Lower-Limb Kinetics and Kinematics. IEEE Sens. J. 2019. [CrossRef] Kim, J.; Bae, M.N.; Lee, K.B.; Hong, S.G. Gait event detection algorithm based on smart insoles. ETRI J. 2019. [CrossRef] Vilas-Boas, M.d.C.; Rocha, A.P.; Choupina, H.M.P.; Cardoso, M.; Fernandes, J.M.; Coelho, T.; Cunha, J.P.S. TTR-FAP Progression Evaluation Based on Gait Analysis Using a Single RGB-D Camera. In Proceedings of the 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Berlin, Germany, 23–27 July 2019; IEEE: Piscataway, NJ, USA, 2019. Cui, X.; Zhao, Z.; Ma, C.; Chen, F.; Liao, H. A Gait Character Analyzing System for Osteoarthritis Pre-diagnosis Using RGB-D Camera and Supervised Classifier. In Proceedings of the World Congress on Medical Physics and Biomedical Engineering, Prague, Czech Republic, 3–8 June 2018; Springer: Berlin, Germany, 2018. Sensors 2020, 20, 957 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 25 of 25 Barnea, A.; Oprisan, C.; Olaru, D. Force Sensitive Resistors Calibration for the Usage in Gripping Devices. Diagn. Predict. Mech. Eng. Szstems DIPRE 2012, 3, 18–27. Parmar, S.; Khodasevych, I.; Troynikov, O. Evaluation of flexible force sensors for pressure monitoring in treatment of chronic venous disorders. Sensors 2017, 17, 1923. [CrossRef] Insole Energy Harvesting from Human Movement Using Piezoelectric Generators. Available online: https://trepo.tuni.fi/handle/123456789/25768 (accessed on 9 February 2020). Development of a Smart Insole System for Real-Time Detection of Temporal Gait Parameters and Related Deviations in Unilateral Lower-Limb Amputees. Available online: https://scholarlyrepository.miami.edu/ cgi/viewcontent.cgi?article=1393&context=oa_theses (accessed on 9 February 2020). Sacco, I.; Amadio, A. A study of biomechanical parameters in gait analysis and sensitive cronaxie of diabetic neuropathic patients. Clin. Biomech. 2000, 15, 196–202. [CrossRef] Menz, H.B.; Lord, S.R.; St George, R.; Fitzpatrick, R.C. Walking stability and sensorimotor function in older people with diabetic peripheral neuropathy1. Arch. Phys. Med. Rehabil. 2004, 85, 245–252. [CrossRef] [PubMed] Dementyev, A.; Hodges, S.; Taylor, S.; Smith, J. Power consumption analysis of Bluetooth Low Energy, ZigBee and ANT sensor nodes in a cyclic sleep scenario. In Proceedings of the 2013 IEEE International Wireless Symposium (IWS), Beijing, China, 14–18 April 2013; IEEE: Piscataway, NJ, USA, 2013. Sacco, I.; Amadio, A. Influence of the diabetic neuropathy on the behavior of electromyographic and sensorial responses in treadmill gait. Clin. Biomech. 2003, 18, 426–434. [CrossRef] Kwon, O.-Y.; Minor, S.D.; Maluf, K.S.; Mueller, M.J. Comparison of muscle activity during walking in subjects with and without diabetic neuropathy. Gait Posture 2003, 18, 105–113. Akashi, P.M.; Sacco, I.C.; Watari, R.; Hennig, E. The effect of diabetic neuropathy and previous foot ulceration in EMG and ground reaction forces during gait. Clin. Biomech. 2008, 23, 584–592. [CrossRef] [PubMed] Sacco, I.C.; Akashi, P.M.; Hennig, E.M. A comparison of lower limb EMG and ground reaction forces between barefoot and shod gait in participants with diabetic neuropathic and healthy controls. BMC Musculoskelet. Disord. 2010, 11, 24. [CrossRef] Gomes, A.A.; Onodera, A.N.; Otuzi, M.E.; Pripas, D.; Mezzarane, R.A.; Sacco, N.; Isabel, C. Electromyography and kinematic changes of gait cycle at different cadences in diabetic neuropathic individuals. Muscle Nerve 2011, 44, 258–268. Sawacha, Z.; Spolaor, F.; Guarneri, G.; Contessa, P.; Carraro, E.; Venturin, A.; Avogaro, A.; Cobelli, C. Abnormal muscle activation during gait in diabetes patients with and without neuropathy. Gait Posture 2012, 35, 101–105. Zhu, H.; Wertsch, J.J.; Harris, G.F.; Loftsgaarden, J.D.; Price, M.B. Foot pressure distribution during walking and shuffling. Arch. Phys. Med. Rehabil. 1991, 72, 390–397. © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).