Human Movements Separation Based on Principle Component Analysis

With more and more attention to terrorist attacks, rescue after disaster and medical treatments, the study on human motions has become a hot topic in recent years. Thanks to the unique mechanism of humans, the m-D signatures, which contain extensive information, of each segment are obviously distinct. It remains a great challenge to separate the movement of humans' each part. In this paper, a method for human movements separation based on a principle component analysis (PCA) is proposed. As one of the classical methods in the blind source separation problems, PCA decomposes the signal to a series of orthogonal basis functions to construct the Eigen subspace. The original signal can be represented by the linear combination of the orthogonal basis functions. In addition, the Akaike information criterion is utilized to determine the minimal number of output for PCA. Furthermore, the ixegram and the optimization theory are combined to cluster the principle components to three new groups. Each group depicts one motion form of human. Simulated results verify the superiority of the proposed algorithm.