作者: Shu-Juan Peng
DOI: 10.1109/CIS.2010.54
关键词:
摘要: The motion segmentation is to divide the original sequence into several fragments with specific semantic, which plays an important role in compression, classification, synthesis. This paper presents a algorithm based on central distance features and low-pass filter for human capture data. proposed approach mainly includes three steps. Firstly, set of from center joint ROOT limbs was extracted, those were divided upper lower norms. Then, PCA method used get one dimension principal component, can better represent motion. Furthermore, utilized denoising signal. Consequently, segmental points be obtained. Experimental results show promising performance our algorithm.