Optimising Real-Time Fall Detection: A Comparative Study of Machine Learning Algorithms Using IMU Sensor

Falls can happen to anyone and require rapid detection to minimise injury risks such as fractures, reduced mobility, and increased dependency. This study proposes a real-time fall detection algorithm using an IMU sensor to capture body movement data (acceleration, angular velocity, and orientation). It compares four machine learning models, Support Vector Machine (SVM), Random Forest (RF), Convolutional Neural Network (CNN), and Artificial Neural Network (ANN) to identify the optimal solution. Data were collected through simulated fall and non-fall activities, including walking, standing, and directional falls, synchronised with IMU readings at 100Hz. ANN achieved the best performance (98% accuracy, 0.1s latency), followed by CNN (97<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), RF (96<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), and SVM (90<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>). Demonstrating its suitability for immediate fall recognition, the system is designed to trigger a protective response (e.g., airbag activation) upon detecting a fall. These results highlight the potential of ANN-based algorithms for reliable and low-latency fall detection in wearable devices.