Neural Network based Realtime Walking Speed Estimation and Gait Phase Detection using Smart Insoles

Three machine learning based gait analysis models have been developed for real-time walking speed estimation, offline gait phase detection and real-time prediction of incoming gait cycle profile. All these have been developed by performing rigorous trials on multiple subjects to make these models deployable for any new subject with features within the extremum considered in model development. These utilize a single sensor based lucid and economical hardware for effective gait analysis in both clinical and outdoor environment. While the offline gait analysis provide insight into characteristics gait patterns of individuals, the real-time estimations and predictions are effective in designing control schemes for wearable assistive devices.