Data in brief 23 (2019) 103839 Contents lists available at ScienceDirect Data in brief journal homepage: www.elsevier.com/locate/dib Data Article A dataset for the development and optimization of fall detection algorithms based on wearable sensors Valentina Cotechini, Alberto Belli, Lorenzo Palma, Micaela Morettini, Laura Burattini*, Paola Pierleoni  Politecnica delle Marche, Ancona, Italy Department of Information Engineering, Universita a r t i c l e i n f o a b s t r a c t Article history: Received 4 February 2019 Received in revised form 1 March 2019 Accepted 7 March 2019 Available online 15 March 2019 This paper describes a dataset acquired on 8 subjects while simulating 13 types of falls and 5 types of Activities of Daily Living (ADL), each repeated 3 times. In details, data includes 4 simulated falls forward (falling on knees ending up lying, ending in lateral position, ending up lying, ending up lying with recovery), 4 backward (falling sitting ending up lying, ending in lateral position, ending up lying, ending up lying with recovery), 2 lateral right (ending up lying, ending up lying with recovery), 2 lateral left (ending up lying, ending up lying with recovery), and 1 syncope. Simulated ADL are: lying on a bed then standing; walking a few meters; sitting on a chair then standing; go up or down three steps; and standing after picking something. Data were acquired using a MARG sensor, a wearable multisensory device tied to the subject's waist, that recorded time-variations of the subject's acceleration and orientation (expressed through the yaw, pitch and roll angles). These data can be useful in the development and test of algorithms to automatically identify and classify fall events. Fall detection systems are particularly useful when a subject is alone and not able to stand up after a fall, since an automatic alarm can be sent remotely to receive proper help. © 2019 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: Human fall Fall detection MARG (Magnetic Angular Rate and Gravity) sensor Wearable device * Corresponding author. E-mail addresses: vcotechini@gmail.com (V. Cotechini), a.belli@univpm.it (A. Belli), l.palma@univpm.it (L. Palma), m.morettini@univpm.it (M. Morettini), l.burattini@univpm.it (L. Burattini), p.pierleoni@univpm.it (P. Pierleoni). https://doi.org/10.1016/j.dib.2019.103839 2352-3409/© 2019 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/). 2 V. Cotechini et al. / Data in brief 23 (2019) 103839 Specifications table Subject area More specific subject area Type of data How data was acquired Bioengineering of movement Fall detection Graph, video Wearable MARG (Magnetic Angular Rate and Gravity) sensor integrating a magnetometer (HMC5883L, Honeywell, USA), an accelerometer (ADXL345, Analog Devices, USA) and a gyroscope (ITG-3200, InvenSense Inc., USA). Data format Raw and analyzed Experimental factors Raw data from sensors and analyzed data obtained from data fusion algorithm. Experimental Healthy young subjects simulating 13 falls (4 forward, 4 backward, 2 lateral right, 2 lateral left, 1 features syncope) and 5 actions of daily living, while wearing MARG devise. Data source location Sensor Network and Internet of Things laboratory, Department of Information Engineering, Universit a Politecnica delle Marche, Ancona, Italy Data accessibility Data is with this article Related research P. Pierleoni, A. Belli, L. Palma, M. Pellegrini, L. Pernini, S. Valenti, A High Reliability Wearable Device for Elderly Fall Detection, IEEE SENSORS JOURNAL VOL. 15 NO. 8 (2015) 4544-53 [1] article Value of the data This kind of data could be useful:  to develop algorithms to automatically identify and classify a fall event;  to test and compare already existing fall-detection algorithms;  to develop alarm systems for falls occurring at home especially among people (like elderly) who might be unable to seek for help;  to support future studies on biomechanics of human fall. 1. Data The proposed dataset is the only one that provides a complete set of data detected through a MARG sensor placed at the waist of the subject representing the most comfortable and least invasive position for wearing a sensor. Moreover, compared to other datasets in the literature [2], it is the only one that presents a complete set of falls including syncope also defined as backwards fall against a wall. Data are organized in two main directories, Fall and ADL each containing several folders as depicted in Fig. 1; each path ends with three folders, FileTXT and Graph (as represented for ADL/Lying&Stand only), containing the data files of each subject, and TrainingVideo, containing files .mp4. The files inside FileTXT are called Si_j.txt, where i ¼ 1,2…8 indicates the subject, and j ¼ 1,2,3 indicates the repetition. Each .txt file is composed of 14 columns, separated by semi column, that represent the timeseries of the following parameters listed in the first row: - 1st column is time in seconds (s); - 2nd to 4th columns contain acceleration along x, y and z axes (acc_x, acc_y, acc_z) expressed in gravitational acceleration g (1 g ¼ 9.80665 m/s2); - 5th to 7th columns contain angular rate with respect to x, y and z axes (gyr_x, gyr_y, gyr_z), expressed in degree per second (⁰/s); - 8th to 10th columns (mag_x, mag_y, mag_z) contain the Earth's magnetic field along x, y and z axes, expressed in Gauss (G), 1 G ¼ 110 4 T (Tesla); - 11th column contains the Signal Vector Magnitude (SVM) of the acceleration in g, computed as the square root of the sum of squared acceleration components; - 12th to 14th columns contain orientational yaw, pitch and roll angles around z, y and x axes, respectively, expressed in degree ( ). V. Cotechini et al. / Data in brief 23 (2019) 103839 3 Fig. 1. Data folders organization. The length of recordings is variable among different subjects, trials and repetitions; thus, the number of rows in each .txt file varies accordingly. The files inside Graph are called Si_j.jpg, where i ¼ 1,2…8 indicates the subject, and j ¼ 1,2,3 indicates the repetition. Each .jpg file contains two subplots; the upper plot represents the SVM of the acceleration over time, while lower plot represents the trend of the pitch and roll angles over time. The files inside TrainingVideo show how to simulate a specific fall or Activities Daily Living (ADL). Overall, there are 432 .txt files, 432 .jpg files and 18 .mp4 files. 2. Experimental design, materials and methods The experiment was performed on 8 healthy volunteers (6 males and 2 females, from 22 to 29 years old) and was carried out in accordance with the Declaration of Helsinki. Each subject signed an informed written consent before participating. The experimental protocol was proposed by Noury et al. [3] and previously used in Refs. [1,4] and partially in Ref. [5] to simulate realistic scenarios of falls and ADL. After a proper video training, each subject has simulated, for 3 times, 13 types of falls and 5 types of ADL as reported in Table 1. Specifically, 4 simulated falls are forward (falling on knees ending up lying, ending in lateral position, ending up lying, ending up lying with recovery), 4 backward (falling sitting ending up lying, ending in lateral position, ending up lying, ending up lying with recovery), 2 lateral right (ending up lying, ending up lying with recovery), 2 lateral left (ending up lying, ending up lying with recovery), and 1 syncope. Simulated ADL are lying on a bed then standing, walking a few meters, sitting on a chair then standing, go up or down three steps, and standing after picking something. Each subject has worn an elastic belt on which was fixed, through adhesive Velcro, the device described in Ref. [1], including a MARG sensor inside a hard-plastic box. The MARG sensor integrates 3axis HMC5883L magnetometer (Honeywell, USA, with resolution of 4 mG in ±8 G fields), ADXL345 accelerometer (Analog Devices, USA, with resolution of 4 mg/LSB in ±16 g range) and ITG-3200 gyroscope (InvenSense Inc., USA, with resolution of 14.375 LSBs per  /s in ±2000 /s range). A proper data fusion algorithm applied to the raw signals of the three sensors [6] was used in Ref. [1] to compute yaw, pitch and roll angles representing the subject's orientation, i.e. subject's rotation around the z, y and x axis, respectively. 4 V. Cotechini et al. / Data in brief 23 (2019) 103839 Table 1 Classification of the simulations performed in the experiment. FALL Forward Backward Lateral right Lateral left Syncope ADL falling on knees ending up lying ending in lateral position ending up lying ending up lying with recovery falling sitting ending up lying ending in lateral position ending up lying ending up lying with recovery ending up lying ending up lying with recovery ending up lying ending up lying with recovery slipping against a wall ending up sitting lying on a bed then standing walking a few meters sitting on a chair then standing climbing three steps standing after picking something Falls were performed safely with the supervision of support staff on a 15 cm thick mattress. Acknowledgments This work funded by the Italian Ministry of Education, University and Research. It was carried out in the Sensor Network and Internet of Things Laboratory by the Telecommunication and Bioengineering  Politecnica delle Marche, Italy. Authors wish to thank the students who Groups at the Universita simulated the falls. Transparency document Transparency document associated with this article can be found in the online version at https:// doi.org/10.1016/j.dib.2019.103839. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.dib.2019.103839. References [1] P. Pierleoni, A. Belli, L. Palma, M. Pellegrini, L. Pernini, S. Valenti, A high reliability wearable device for elderly fall detection, IEEE Sens. J. 15 (2015) 4544e4553. https://doi.org/10.1109/JSEN.2015.2423562.  n, J.M. Cano-García, Analysis of public datasets for wearable fall detection systems, Sensors 17 [2] E. Casilari, J.A. Santoyo-Ramo (2017) 1513e1541. https://doi.org/10.3390/s17071513.  Laighin, V. Rialle, J.E. Lundy, Fall detection - principles and methods, in: [3] N. Noury, A. Fleury, P. Rumeau, A.K. Bourke, G. O Proc. IEEE Engineering in Medicine and Biology Conference, EMBC), 2007, pp. 1663e1666. https://doi.org/10.1109/IEMBS. 2007.4352627. [4] P. Pierleoni, A. Belli, L. Palma, L. Pernini, S. Valenti, A versatile ankle-mounted fall detection device based on attitude heading systems, in: Proc. IEEE Biomedical Circuits and Systems Conference (BioCAS), 2014, pp. 153e156. https://doi.org/10. 1109/BioCAS.2014.6981668. [5] P. Pierleoni, A. Belli, L. Maurizi, L. Palma, L. Pernini, M. Paniccia, S. Valenti, A wearable fall detector for elderly people based on AHRS and barometric sensor, IEEE Sens. J. 16 (2016) 6733e6744. https://doi.org/10.1109/JSEN.2016.2585667. [6] P. Pierleoni, A. Belli, L. Palma, L. Pernini, S. Valenti, An accurate device for real-time altitude estimation using data fusion algorithms, in: Proc. IEEE Mechatronic and Embedded Systems and Applications (MESA), 2014, pp. 1e5. https://doi.org/10. 1109/MESA.2014.6935583.