A Time Reversal Symmetry Based Real-time Optical Motion Capture Missing Marker Recovery Method

This paper proposes a deep learning model based on time reversal symmetry for real-time recovery of continuous missing marker sequences in optical motion capture. This paper firstly uses time reversal symmetry of human motion as a constraint of the model. BiLSTM is used to describe the constraint and extract the bidirectional spatiotemporal features. This paper proposes a weight position loss function for model training, which describes the effect of different joints on the pose. Compared with the existing methods, the experimental results show that the proposed method has higher accuracy and good real-time performance.