Fall Risk Prediction System Using Wearable Sensors and LSTM - Based Machine Learning

• Author(s): Arshiya Fatima ; Shaistha Khanum ; Siddhi Singh Rathor ; Syed Rayyan Ahmed ; Taufiq Ahmed I
• Paper ID: 1708547
• Page: 1270-1277
• Published Date: 21-05-2025
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 8 Issue 11 May-2025

Abstract

Falls represent a significant health hazard, particularly among the elderly and patients with mobility impairments. Timely prediction of fall risk can play a vital role in reducing injuries, hospitalization, and related healthcare costs. This paper proposes a cost-effective and intelligent fall risk prediction system that combines wearable sensors with machine learning to assess the likelihood of falls in real time. The system architecture includes an ESP- 32 microcontroller, an accelerometer sensor (MPU6050), a force pressure sensor, resistors for signal conditioning, a buzzer for immediate alerts, and a common power supply for mobility. These components are embedded into a wearable module that continuously collects motion and pressure data. The collected data is transmitted to a remote server where it is preprocessed and analyzed using a Long Short- Term Memory (LSTM) model trained to detect irregular movement patterns that typically precede a fall. The system demonstrated robust performance, achieving a PSNR value of 10.3 dB and an SSIM score of 0.4, validating its ability to preserve structural features in time-series sensor signals. Performance testing revealed that even with limited training (three epochs due to hardware constraints), the model was capable of identifying motion anomalies associated with fall risk. A buzzer-based local alert mechanism provides instant feedback to the user or nearby caregiver. The usability of the system was positively received, especially for its lightweight design and intuitive functionality. The device holds strong potential for further development, including mobile application integration for cloud-based visualization, enhanced sensor fusion (e.g., gyroscope, ECG), and edge-AI implementation for faster local processing. This research contributes toward creating an affordable, scalable, and effective fall risk prediction solution tailored for home-based elderly care, post-operative rehabilitation, and assistive living facilities.

Keywords

Fall Risk Prediction, LSTM, Wearable Sensors, ESP-32, Accelerometer, Force Pressure Sensor, IoT in Healthcare, MIT App Inventor, Real-Time Monitoring, Elderly Care

Citations

IRE Journals:
Arshiya Fatima , Shaistha Khanum , Siddhi Singh Rathor , Syed Rayyan Ahmed , Taufiq Ahmed I "Fall Risk Prediction System Using Wearable Sensors and LSTM - Based Machine Learning" Iconic Research And Engineering Journals Volume 8 Issue 11 2025 Page 1270-1277

IEEE:
Arshiya Fatima , Shaistha Khanum , Siddhi Singh Rathor , Syed Rayyan Ahmed , Taufiq Ahmed I "Fall Risk Prediction System Using Wearable Sensors and LSTM - Based Machine Learning" Iconic Research And Engineering Journals, 8(11)