Volume 9, Issue 12, December – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Fall Detection System for Elderly Using

IoT Technology
Mandeep Singh Punnay Raheja
Computer Science and Engineering Computer Science and Engineering
Manav Rachna International Institute of Research and Studies Manav Rachna International Institute of Research and Studies
Faridabad, Haryana, India Faridabad, Haryana, India

Shweta Sharma Lakshay Sharma

Computer Science and Engineering Computer Science and Engineering
Manav Rachna International Institute of Research and Studies Manav Rachna International Institute of Research and Studies
Faridabad, Haryana, India Faridabad, Haryana, India

Abstract:- Linking every aspect of our lives could have Death may also occur if emergency medical care is not
immediate positive effects on society. A basic gadget can received right away.
be included in the phrase "Internet of Things" if we give
it "computational intelligence" and connect it to the The fall detection system is quite valuable for the
network. In addition, improving features of the elderly. It can inform the appropriate individual or family
fundamental design, the "smart" gadget is typically member if it senses a fall, reducing the risk of delayed
portable and a option that is more affordable, effective, medical assistance. This has resulted in the creation of
and has the potential to grow in functionality over time. numerous types of automated fall detection systems.
IoT is changing our houses to better meet each person's Smartwatches, fitness trackers, and other wearable devices
needs and desires. now include fall detectors.

The goal of our IoT-based fall detection system II. LITERATURE REVIEW
project for smart home environments is still similar to
this, but it has more room to grow in terms of usefulness. In [1] the authors discussed the progression of fall
Not only would this gadget sound an alarm in the event detection techniques utilizing skeleton data derived from
that an elderly person sustains injuries from falls but can RGB videos. It discusses both conventional handcrafted
also be applied to identify costly things that fall when methods and contemporary deep learning (DL) strategies.
being similar to stores that keep opulent and high-end [ Handcrafted techniques generally focus on the extraction of
merchandise on display for customers. Additionally, this features such as joint coordinates, angles, and distances from
prototype can be incorporated to learn popular the skeleton data, whereas DL approaches utilize neural
interaction models. nowadays days in IoT devices, like networks to autonomously identify spatio-temporal patterns.
video monitoring and voice help. The benefits of skeleton-based methods include their non-
intrusive characteristics and reduced costs in comparison to
Keywords:- IoT, Fall Detection, Video Monitoring, Sensor wearable and ambient sensor-based alternatives. Nonetheless,
Integration. challenges such as background interference in RGB videos
and the necessity for sophisticated models present limitations.
I. INTRODUCTION The results underscore the efficacy of deep learning,
especially with architectures like Graph Convolutional
Elderly individuals who live alone run the danger of Networks (GCNs) and Transformers, in enhancing accuracy.
having mishaps at home. According to a prior case study, the Future research directions stress the importance of
majority of elderly patients are admitted to hospitals due to developing robust models capable of functioning in real-time
falls carrying out their regular tasks. According to World and across varied environments.
Health Organization (WHO), 28% to 35% of older adults that
were of the age range 65 and older experienced fall each year, In [2] the authors present a novel approach to fall
with the percentage rising to 32% to 42% for people who are detection by utilizing spatio-temporal features derived from
older than 70. Falls may have an impact on the health of RGB video frames. This technique involves measuring
elderly people. Given that they As they get older, their distances and angles between key points of an individual's
physical strength declines. In addition, autumn occurrences skeleton across consecutive frames, identified through a 2D
could result in catastrophic injuries such bone fractures and model. To enhance the analysis, Principal Component
traumas including paralysis and brain failure. Analysis (PCA) is applied to minimize the dimensionality of
the extracted features. Various classifiers, including Support

IJISRT24DEC108 www.ijisrt.com 210

Volume 9, Issue 12, December – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Vector Machine (SVM), Decision Tree, Random Forest, and Espressif Systems developed the ESP8266, a low-cost Wi-Fi
K-Nearest Neighbors (KNN), are utilized to differentiate chip that supports the TCP/IP protocol.[5]
between fall and non-fall incidents. The method boasts
several advantages, such as being non-intrusive and  The Esp8266 Chip
achieving a high detection accuracy of 98.5%, with a The ESP8266 chip features a 32-bit LX106 RISC CPU
sensitivity of 97% and a specificity of 100%. Nonetheless, it with customizable clock frequency of 80 to 160 MHz and
is important to note that the effectiveness of this approach is support for RTOS. The device features 128kb of RAM and
influenced by lighting conditions and the positioning of 4mb of flash memory, allowing for quick connectivity and
cameras. The results indicate that SVM is the most effective operation. It integrates an 802.11b/g/n HT40 Wi-Fi
classifier for fall event detection, highlighting its significant transceiver that connects to the network and allows for self-
potential for real-time implementation in systems designed configuration. This makes the chip incredibly adaptable. It
for elderly care. will connect based on the wifi protocol setup.

In [3] the authors introduce a fall detection system that

utilizes millimeter-wave radar to analyze one-dimensional
point clouds and Doppler velocity information. This
approach incorporates Long Short-Term Memory (LSTM)
networks for real-time classification of falls. Notable benefits
of this system include the preservation of privacy, as radar
technology can operate effectively through walls and under
various lighting conditions, rendering it adaptable to multiple
settings. However, there are limitations, such as the
possibility of false alarms stemming from erroneous data,
which are addressed through preprocessing methods like
sliding window processing and filtering. The results indicate
that the system can achieve a high accuracy rate of up to 99.5%
while maintaining low computational complexity, thereby
proving to be efficient for real-time fall detection in the
elderly.

In [4] the authors present a fall detection mechanism Fig 1: NodeMcu Development Board Kit
utilizing a wrist-mounted device, the M5StickC Plus, which
gathers acceleration data via its 6-axis accelerometer. This  Mpu6050
system employs a streamlined algorithm that relies on It is a 6-axis motion sensing device. In this device a 3-
statistical analysis of acceleration variations within a one- axis gyroscope and 3-axis accelerometer work together to
second timeframe to identify falls. It specifically examines detect motions like falling. It is compatible with NodeMcu
the "walk-fall-still" sequence, activating an alarm in real-time and can work for both fast and slow motions. Fig. 2 shows a
upon detecting a fall. The system's benefits include a rapid MPU6050 module with its pin out diagram.
response time of one second, minimal computational
demands, and a non-intrusive design, making it particularly
advantageous for applications in elderly care. Nonetheless, it
faces challenges such as generating false positives during
specific hand movements, such as hand-fall or hand-wave
actions, and its battery life is constrained by the limited
capacity of the wrist-worn device. The research indicates that
the system achieves an accuracy rate of 90%, and while it is
not flawless, it demonstrates considerable promise for
implementation in effective fall detection solutions.

III. SYSTEM SETUP

A. Hardware Components
Our fall detection was built utilizing a variety of parts,
including communication modules and sensors. Important
elements include:
Fig. 2: MPU6050
 Nodemcu
It is an open-source platform, with the hardware design  3-Axis Gyroscope
available for editing, modification, and building. The It measures rotational velocity (rad/s) along the X, Y,
NodeMCU Dev Kit/board has an ESP8266 Wi-Fi chip. and Z-axis. This is called angular velocity, which is the rate
of change of angular position over time. The gyroscope uses

IJISRT24DEC108 www.ijisrt.com 211

Volume 9, Issue 12, December – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

micro electro-mechanical system (MEMS) technology to IV. METHODOLOGY

detect vibrations caused by the Coriolis effect when the
sensor is rotated along any axis. The MPU-6050 has a 16-bit A. Level 1: - Hardware Setup
analog-to-digital converter (ADC) that digitizes the voltage The essential elements utilized in this project consist of
to sample each axis, and the sampling rate can be adjusted the ESP32 microcontroller and the MPU6050 sensor, which
from 3.9 to 8000 samples per second. The gyroscope has four combines an accelerometer and gyroscope to facilitate real-
programmable full-scale ranges of ±250°/s, ±500°/s, time motion tracking. The connection between the MPU6050
±1000°/s, and ±2000°/s, and angular velocity is measured in and the ESP32 is established through the I2C communication
degrees per second.[7][8] protocol. The choice of the ESP32, a cost-effective
microcontroller equipped with Wi-Fi capabilities, is
 3-Axis Accelerometer: attributed to its adaptability and seamless integration with
It works by detecting acceleration on each coordinate Internet of Things (IoT) applications. The MPU6050 sensor
axis. For example, if the MPU6050 is stationary on a desktop, plays a vital role in assessing orientation, movement, and
the Z-axis acceleration will be 1 gravity unit, while the X and acceleration. These hardware components were deliberately
Y axes will be 0. If the MPU6050 is tilted or in a weightless selected to ensure that the fall detection system is both
or overweight condition, the corresponding reading will portable and dependable, making it appropriate for both
change.[7][8] residential and industrial environments. The assembly of
these components was executed on a breadboard, with careful
B. Software Components attention to the connections to guarantee that data from the
These are the software components of our system: - MPU6050 is transmitted accurately to the ESP32. This
configuration enables the system to identify abrupt variations
 Arduino IDE in movement or acceleration, which are characteristic of a
Arduino IDE, or Integrated Development Environment, fall.
is an open-source software that allows users to write and
upload code to an Arduino board. It's available for download B. Level 2: - Fall Detection Algorithm
on Windows, Linux, and macOS, and is compatible with At the heart of the fall detection system lies a threshold-
many operating systems. The Arduino IDE shown in, based algorithm that analyses real-time data obtained from
includes a C++ code editor, the GNU C++ compiler, and a the MPU6050 sensor. This sensor's accelerometer
program upload utility. Users can create new projects, open continuously tracks the user's movements, focusing
existing examples, and edit, compile, and upload code to particularly on variations in acceleration and orientation
their Arduino device. Programs written using the Arduino across the three axes: X, Y, and Z. The algorithm is
IDE are called sketches and are saved with the file extension specifically engineered to identify a fall when the
‘.ino’.[9][10] acceleration surpasses a predetermined threshold. For
example, abrupt declines in the Z-axis or significant spikes in
 IFTTT (If This Then That): acceleration can signal a fall event. This approach guarantees
It is a free web-based service that allows users to accurate fall detection while reducing the likelihood of false
automate tasks between various apps and devices by alarms that may arise from ordinary movements.
connecting them with simple command chains called applets. Furthermore, the algorithm is fine-tuned to distinguish
Applets are made up of two parts: a trigger and an action. between typical activities such as walking or sitting and
The trigger is the condition or event that activates the applet, genuine falls, thereby improving the system's overall
and the action is the result that occurs when the trigger is sensitivity and specificity.
activated.[11][12]

IJISRT24DEC108 www.ijisrt.com 212

Volume 9, Issue 12, December – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig 3: Algorithm of fall detection system

C. Level 3: - Alert Mechanism consistently monitor the sensor data, employing an algorithm
Upon detection of a fall, the ESP32 promptly issues an to detect falls and trigger alerts as necessary. Utilizing IFTTT
alert through email via the IFTTT (If This Then That) streamlines the alerting process, providing flexibility for
platform. Leveraging the Wi-Fi capabilities of the ESP32, future modifications, enhancements, or the integration of
this system facilitates smooth integration with online services. additional notification methods such as SMS or app alerts.
It initiates an HTTP request directed to IFTTT’s webhook The programming is executed in a blend of C and C++,
service, which is set up to dispatch an email notification to a ensuring real-time processing of the sensor data.
designated recipient. This process guarantees immediate
alerting, enabling caregivers or appropriate authorities to act V. RESULT
swiftly in response to fall incidents. The email provides
pertinent information regarding the fall, such as the time and The fall detection algorithm serves as the fundamental
the precise moment the threshold was breached, thereby element of the system, tasked with recognizing potential fall
serving as an effective communication tool for monitoring incidents and activating the alert mechanism. The process
elderly individuals or fragile items during transit. commences with the establishment of a reliable hardware
configuration, which involves initializing communication
D. Level 4: - Software between the ESP32 microcontroller and the MPU6050 sensor
The development of the ESP32 in conjunction with the through the I2C protocol. This connection allows for the
MPU6050 sensor is accomplished through the Arduino IDE. efficient transfer of acceleration and gyroscope data from the
The implemented code incorporates libraries that facilitate sensor to the microcontroller for immediate analysis. To
the I2C communication with the MPU6050 and enable the guarantee the precision of the measurements, the system
processing of both accelerometer and gyroscope data. The undertakes a thorough calibration of the MPU6050 sensor,
ESP32 is set up with the necessary Wi-Fi credentials to ensuring accurate readings along the X, Y, and Z axes.
establish communication with IFTTT, allowing for the Additionally, the ESP32 is set up to connect to a local Wi-Fi
transmission of alerts. The software is designed to network by entering the required SSID and password. This

IJISRT24DEC108 www.ijisrt.com 213

Volume 9, Issue 12, December – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Wi-Fi connectivity enables interaction with the IFTTT elderly falls, we achieved a detection accuracy of 97.2%, a
platform or other notification services, thereby facilitating the response time of 2.1 seconds, a sensitivity of 96.7%, and a
prompt transmission of alerts upon the detection of a fall. specificity of 97.5%. When an elderly person tripped, the
system had a detection accuracy of 95.8%, a response time of
To evaluate the performance of our IoT-based fall 2.5 seconds, a sensitivity of 94.3%, and a specificity of 96.7%.
detection system, we conducted a series of experiments in a
simulated home environment. The system was tested under The system also performed well in detecting falls of
various scenarios, including simulated elderly falls, elderly objects, with an accuracy of 92.4%, a response time of 3.1
persons tripping, objects falling, and everyday movements. seconds, a sensitivity of 91.2%, and a specificity of 93.6%.
The accuracy, response time, sensitivity, and specificity were Even for everyday movements, the system maintained a high
measured for each test case. detection accuracy of 98.1%, a response time of 1.8 seconds, a
sensitivity of 97.9%, and a specificity of 98.4%.
The results, shown in the given table, demonstrate the
effectiveness of our fall detection system. For simulated

Table 1: Result

These results indicate that our IoT-based fall detection VII. CONCLUSION
system is capable of accurately identifying falls, while
minimizing false alarms and providing timely alerts. The The IoT-based fall detection system signifies a notable
system's ability to differentiate between different types of progression in improving the safety and welfare of elderly
events, such as falls of elderly individuals and falling objects, individuals and those susceptible to falls. The effective
highlights its robustness and potential for practical amalgamation of sensors, including the MPU6050
applications in smart home environments. accelerometer and gyroscope, in conjunction with the ESP32
microcontroller and IFTTT platform, underscores the
VI. DISCUSSION transformative potential of these technologies in fall
monitoring and emergency response. This system illustrates
The fall detection system presented in this research the vital role of prompt fall detection in minimizing response
highlights the remarkable functionality of the MPU6050 times and enhancing overall safety through the strategic
sensor in conjunction with the ESP32 microcontroller for the placement of motion sensors. With its versatile software
reliable identification of fall incidents. Utilizing a threshold- algorithms and wireless communication features, the system
based algorithm, the system continuously evaluates real-time has demonstrated the importance of real-time monitoring and
data from the accelerometer and gyroscope, allowing it to instant alert notifications, facilitating timely intervention and
recognize abrupt shifts in movement and orientation that care for individuals in need. The incorporation of Bluetooth
signify a fall. This methodology not only guarantees precise control and IFTTT integration further augments the system's
fall detection but also significantly reduces the chances of adaptability, permitting future enhancements and the
false alarms that could result from ordinary daily activities. inclusion of additional notification methods, such as SMS or
mobile application alerts.
The alert system, which capitalizes on the ESP32's Wi-
Fi capabilities, integrates smoothly with the IFTTT platform FUTURE WORK
to swiftly send email notifications to pre-selected caregivers
or emergency contacts upon detecting a fall. These Future investigations may focus on the incorporation of
notifications include essential information such as the time of artificial intelligence (AI) to enhance real-time fall detection
the incident and the specific moment the threshold was and prediction capabilities. The creation of AI-based
exceeded, facilitating a prompt and informed response to algorithms could enable systems to more effectively
emergencies. The software development, which employs a differentiate between falls and non-fall activities, thereby
combination of C and C++, ensures the real-time processing minimizing false positives and increasing overall reliability.
of sensor data and efficient integration with online services, Furthermore, machine learning models can be developed to
rendering the fall detection system a reliable and effective manage intricate movement patterns, which would further
solution for safeguarding the well-being of elderly elevate detection accuracy across various environments and
individuals or fragile items during transportation. demographic groups, such as the elderly or those with
physical impairments.[13][14]

IJISRT24DEC108 www.ijisrt.com 214

Volume 9, Issue 12, December – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

One additional promising domain is edge computing, [11]. S. Vorapojpisut, "A lightweight framework of home
which facilitates real-time data processing directly on the automation systems based on the IFTTT model,"
device. This approach reduces latency and improves privacy Journal of Software, vol. 10, no. 12, pp. 1343-1350,
by lessening reliance on cloud-based data processing. By 2015. https://doi.org/10.17706/jsw.10.12.1343-1350
integrating edge computing, fall detection systems can [12]. V. Kanade, "Understanding the meaning of IFTTT, its
operate effectively in remote regions or areas with restricted working, and alternatives," Spiceworks Inc., Aug. 31,
connectivity. [15] 2023. [Online]. Available:
https://www.spiceworks.com/tech/tech-
Further investigations should prioritize the advancement general/articles/what-is-ifttt/
of more durable and user-friendly sensors that minimize [13]. M. A. Sarwar, B. Chea, M. Widjaja, and W. Saadeh, "An
intrusiveness for individuals. The creation of lightweight and AI-Based Approach for Accurate Fall Detection and
comfortable devices that can be effortlessly incorporated into Prediction Using Wearable Sensors," in 2024 IEEE 67th
everyday attire would enhance user adherence, especially International Midwest Symposium on Circuits and
among older adults, who represent the main demographic for Systems (MWSCAS), Springfield, MA, USA, 2024, pp.
these systems. [16] 118–121. doi: 10.1109/MWSCAS60917.2024.10658849.
[14]. D. Giuffrida, G. Benetti, D. De Martini, and T.
REFERENCES Facchinetti, "Fall Detection with Supervised Machine
Learning Using Wearable Sensors," in 2019 IEEE 17th
[1]. V. Hoang, J. W. Lee, M. J. Piran, and C. Park, International Conference on Industrial Informatics
"Advances in Skeleton-Based Fall Detection in RGB (INDIN), Helsinki, Finland, 2019, pp. 253–259. doi:
Videos: From Handcrafted to Deep Learning 10.1109/INDIN41052.2019.8972246.
Approaches," IEEE Access, vol. 11, pp. 92322–92352, [15]. B. -S. Lin et al., "Fall Detection System with Artificial
2023. https://doi.org/10.1109/access.2023.3307138 Intelligence-Based Edge Computing," IEEE Access, vol.
[2]. A. Y. Alaoui, S. El Fkihi, and R. O. H. Thami, "Fall 10, pp. 4328–4339, 2022. doi:
Detection for Elderly People Using the Variation of Key 10.1109/ACCESS.2021.3140164.
Points of Human Skeleton," IEEE Access, vol. 7, pp. [16]. A. Aldousari, M. Alotaibi, F. Khajah, A. Jaafar, M.
154786-154795,2019. doi: Alshebli, and H. Kanj, "A Wearable IoT-Based
10.1109/ACCESS.2019.2946522. Healthcare Monitoring System for Elderly People," in
[3]. C. Kittiyanpunya, P. Chomdee, A. Boonpoonga, and D. 2023 5th International Conference on Bioengineering
Torrungrueng, "Millimeter-Wave Radar-Based Elderly for Smart Technologies (BioSMART), Paris, France,
Fall Detection Fed by One-Dimensional Point Cloud and 2023, pp. 1–4. doi:
Doppler," IEEE Access, vol. 11, pp. 76269-76283, 2023. 10.1109/BioSMART58455.2023.10162041.
doi: 10.1109/ACCESS.2023.3297512.
[4]. S. Li, "Fall Detection With Wrist-Worn Watch by
Observations in Statistics of Acceleration," IEEE
Access, vol. 11, pp. 19567-19578, 2023, doi:
10.1109/ACCESS.2023.3249191.
[5]. Espressif Systems, "ESP8266 Overview." [Online].
Available:
https://www.espressif.com/en/products/hardware/esp82
66ex/overview
[6]. SunFounder, "MPU6050 Documentation," [Online].
Available:
https://docs.sunfounder.com/projects/ultimate-sensor-
kit/en/latest/components_basic/05-
component_mpu6050.html
[7]. P. Robots, "The MPU6050 Explained," Programming
Robots. [Online]. Available:
https://mjwhite8119.github.io/Robots/mpu6050
[8]. S. Santos & R. Santos, "Arduino Guide for MPU-6050
Accelerometer and Gyroscope," Random Nerd Tutorials.
[Online]. Available:
https://randomnerdtutorials.com/arduino-mpu-6050-
accelerometer-gyroscope
[9]. M. Fezari & A. Al Dahoud, "Integrated Development
Environment (IDE) for Arduino," 2018.
[10]. Arduino, "Software," [Online]. Available:
https://www.arduino.cc/en/software

IJISRT24DEC108 www.ijisrt.com 215