3D Human Motion Key-Frames Extraction Based on Asynchronous Learning Factor PSO

Key-frames extraction technology for motion capture data can extract some important frames which describe the original motion sequence well, it is useful in motion retrieval, compression and edition. In this paper, we propose a new method for extracting key frames from motion capture sequence based on asynchronous learning factor PSO (Particle Swarm Optimization). Our proposed approach consists of three steps. Firstly, we initialize every particle which represents a series of key-frames. Secondly, the algorithm searches the global optimal value through asynchronous learning factor. Finally, we get the best key-frames set based on fitness function which is calculated by compression ratio and reconstruction error rate. Experiment results show that our method extracts key-frames efficiently which can describe the original motion sequence well.