A low-cost machine learning process for gait measurement using biomechanical sensors

Abstract Continuous gait measurement can bring relevant indicators for healthcare professionals. Several techniques were developed for this cause. However, the beneficiaries, especially senior adults, find it hard to accept a monitoring device as it takes away their privacy. In this paper, we present a non-intrusive, low-cost and easy to implement model for gait measurement at home. It consists of implementing 4 passive infrared (PIR) sensors facing each other by pair. Our approach is based on a Deep Learning (DL) model that takes as input the signals generated by the PIR sensors, as they are representative of the distance and the speed of the moving object. A temporary Depth camera is used for training the model on the gait parameters. To evaluate our approach, we conducted multiple series of experiments on real sensor data. The results are promising and show that our approach is efficient for continuous gait measurement.