INTRODUCTION :

Fall detection is a critical area of research, especially in fields such as healthcare, elderly care, and workplace safety. The implications of falls are profound, particularly for older adults. [1] Falls can lead to injuries such as fractures, head trauma, and even mortality. In addition, they can result in a fear of falling, which may further limit mobility and independence. [2][3] Studies have shown that individuals who experience falls often suffer from a decline in physical and mental health, leading to increased healthcare costs and reliance on assistance.[4] Therefore, implementing effective fall detection systems is essential for enhancing safety and well-being. Early detection of falls allows for prompt assistance, potentially preventing serious consequences. [5] With the advancement of wearable technology and machine learning algorithms, automated fall detection systems have become increasingly feasible. [6] These systems not only improve response times but also reduce the need for constant monitoring by caregivers or medical personnel.

The methods used for fall detection encompass a wide range, from sensor-based systems to learning approaches. Inertial sensors, such as accelerometers and gyroscopes, are commonly utilized to monitor user movements and detect fall scenarios. Pressure sensors can create early warning mechanisms by detecting changes in pressure on the ground during a fall event.[7] Vision-based systems detect falls by analyzing movement through cameras.[8] Additionally, machine learning and deep learning algorithms enhance the accuracy of fall detection by analyzing large datasets to identify patterns and distinguish between fall and non-fall situations. [9]

These advanced algorithms can improve the system's performance over time as they learn from new data, making them particularly effective in diverse environments and conditions.[10] Furthermore, Doppler radar systems can monitor the motion of individuals in real-time, providing additional data points for detecting falls based on changes in velocity.[11] The effectiveness of these systems is directly related to the diversity and integration of the technologies used. The combination of advanced sensor technologies and machine learning methods facilitates the development of more accurate and rapid fall detection systems.[12] Despite the advancements in technology, several challenges remain in developing effective fall detection systems. Issues such as false positives, where non-fall activities are mistakenly classified as falls, can undermine user trust and system reliability. [13]Additionally, varying mobility patterns among individuals necessitate adaptable algorithms that can learn and differentiate between typical and atypical behaviors. For this, you need a wide variety of data, and collecting and labeling such a large amount of data may not be feasible in terms of time. In this study, we trained and deployed a YOLO fall detection model to speed up and automate the labeling process, and the data was labeled automatically by referencing this model. Additionally, as stated in [14], we used GANs to increase the amount of fall-labeled data. By generating synthetic samples resembling real IMU signals associated with falls, we enhanced the dataset's diversity, which allowed for better generalization of the model and improved fall detection accuracy. This data augmentation technique also addressed the class imbalance problem, ensuring that the model did not overfit to the non-fall events. For time series classification, we applied conventional methods like LSTM, Conv1D and one of the latest models for time series prediction is TSMixer. Additionally, we transformed the data into a spectrogram and performed classification using image-based methods such as Conv2D.

V. REFERENCES

[1] Pech M, Sauzeon H, Yebda T, Benois-Pineau J, Amieva H. Falls Detection and Prevention Systems in Home Care for Older Adults: Myth or Reality? JMIR Aging. 2021 Dec 9;4(4):e29744. doi: 10.2196/29744. PMID: 34889755; PMCID: PMC8704100. [2] Vaishya R, Vaish A. Falls in Older Adults are Serious. Indian J Orthop. 2020 Jan 24;54(1):69-74. doi: 10.1007/s43465-019-00037-x. PMID: 32257019; PMCID: PMC7093636. [3] Heeb MA. Dangers of Falling. Mo Med. 2017 Jul-Aug;114(4):261. PMID: 30228605; PMCID: PMC6140081. [4] LeLaurin JH, Shorr RI. Preventing Falls in Hospitalized Patients: State of the Science. Clin Geriatr Med. 2019 May;35(2):273-283. doi: 10.1016/j.cger.2019.01.007. Epub 2019 Mar 1. PMID: 30929888; PMCID: PMC6446937. [5] Dariusz Mrozek, Anna Koczur, Bożena Małysiak-Mrozek,Fall detection in older adults with mobile IoT devices and machine learning in the cloud and on the edge,Information Sciences,Volume 537,2020,Pages 132-147 [6] Nur Izdihar Muhd Amir, Rudzidatul Akmam Dziyauddin, Norliza Mohamed, et al. Fall Detection System using Wearable Sensor Devices and Machine Learning: A Review. TechRxiv. March 19, 2024. [7] Wearable Fall Detection System Using Barometric Pressure Sensors and Machine Learning [8] Gutiérrez J, Rodríguez V, Martin S. Comprehensive Review of Vision-Based Fall Detection Systems. Sensors (Basel). 2021 Feb 1;21(3):947. doi: 10.3390/s21030947. [9] Usmani S, Saboor A, Haris M, Khan MA, Park H. Latest Research Trends in Fall Detection and Prevention Using Machine Learning: A Systematic Review. Sensors (Basel). 2021 Jul 29;21(15):5134. doi: 10.3390/s21155134. [10] Yelne S, Chaudhary M, Dod K, Sayyad A, Sharma R. Harnessing the Power of AI: A Comprehensive Review of Its Impact and Challenges in Nursing Science and Healthcare. Cureus. 2023 Nov 22;15(11):e49252. doi: 10.7759/cureus.49252. [11] Li, Z.; Du, J.; Zhu, B.; Greenwald, S.E.; Xu, L.; Yao, Y.; Bao, N. Doppler Radar Sensor-Based Fall Detection Using a Convolutional Bidirectional Long Short-Term Memory Model. Sensors 2024, 24, 5365. https://doi.org/10.3390/s24165365 [12] Pech M, Sauzeon H, Yebda T, Benois-Pineau J, Amieva H. Falls Detection and Prevention Systems in Home Care for Older Adults: Myth or Reality? JMIR Aging. 2021 Dec 9;4(4):e29744. doi: 10.2196/29744. [13] Igual R, Medrano C, Plaza I. Challenges, issues and trends in fall detection systems. Biomed Eng Online. 2013 Jul 6;12:66. doi: 10.1186/1475-925X-12-66. [14] Mohammadzadeh, M., Ghadami, A., Taheri, A., & Behzadipour, S. (n.d.). cGAN-Based High Dimensional IMU Sensor Data Generation for Enhanced Human Activity Recognition in Therapeutic Activities.