Wearable Sensor and Machine Learning Model-Based Fall Detection System for Safety of Elders and Movement Disorders

Due to different working culture of people, elderly people's health gets neglected as they live alone at home. With the rising population, there is a pressing demand for the evolution of fall identification systems. People with age greater than 65 are suffering from highest number of fatal falls. Some of the difficulties and challenges faced by the elders and mobility disordered people can be over passed by implementing algorithms able to anticipate falls. It is possible to have a good to great quality of life for the affected, by providing living assistance through automatic fall detection and alarming systems. We propose implementation of a fall recognition system for real-time tracking of elderly people. The proposed system has wearable sensor unit for detecting falls and alert mechanism to intimate the concerned and the care takers in case of falls by means of messages. The acceleration data are collected by the system using triaxial accelerometer and use machine learning algorithms to detect the falls upon various feature calculations. Extensive computations are carried out to compare the performance of different machine learning algorithms with varying features, and the algorithm giving the highest accuracy with optimal features is identified. The system gains an accuracy up to 99% by using random forest algorithm with tenfold cross validation. Thus, with a secure and reliable fall detection and alarming system, one could reduce the fatal falls, improving social integration, productivity, and quality of life.