Motion Reconstruction and Simulation Using Sparse Inertial Sensors

A method was proposed to reconstruct high-dimensional full-body motion sequences from low-dimensional control data collected by sparse inertial sensors. The approach solved the mapping problem from low dimension to high dimension. A numerical similarity-geometrical similarity-time continuity model was setup to ensure the reconstructed motion candidates in numerical-logical similarity. The gap between angle and angular acceleration was eliminated by acceleration reconstruction. An energy function was introduced to optimize the reconstructed results which guaranteed the accuracy. The analysis and comparison experiments show that the proposed method can reconstruct nature and credible motions in real-time and can be applied in low-cost full-body motion capture by using few inertial sensors.