Real-time body orientation estimation based on two-layer stochastic filter architecture

Abstract

The article presents real time rigid body orientation estimation using inertial and magnetic sensors. Based on the review of orientation estimation literature we suggest, as possible alternative to standard approaches, novel two-layer stochastic estimation filter architecture based on Kalman and particle filters combined into two layers. Two-layer architecture was chosen because it enables greater applicability via upgrade of already implemented Kalman or particle filters. Four two-layer filter architectures were designed, each one enabling different layer interaction. Estimation of human head orientation was chosen as a case example. Simulated data and batch head orientation measurement data were used to test the proposed architectures in terms of accuracy and computational efficiency and to select the best one in terms of aforementioned performance parameters. Selected architecture was then implemented in real time for human-computer interaction and was tested on several practical applications. Obtained results are presented and discussed and future research directions suggested.