Systems and methods for real-time tracking of trajectories using motion sensors

Abstract

Tracking motion using inertial sensors embedded in commercial grade wearables like smartwatches has proved to be a challenging task, especially if real-time tracking is a requirement. Present disclosure provides system and method wherein data from sensors are obtained and scaled. Further, Euler Rodrigues Matrix (ERM) is generated based delta value obtained using sensor data. The scaled sensor data and ERM are used for generating feature vectors. Windowing technique is applied for subsets of feature vectors to obtain label for each window and machine learning model is trained with the label and window. Further, during real-time, sensor data is obtained, and steps of ERM, feature vectors generation, and application of windowing technique are repeated, and coordinates are estimated for each window in real-time based on which trajectories are tracked in real-time for each window.