Hierarchical Aligned Cluster Analysis ( HACA ) for Temporal Segmentation of Human Motion

Temporal segmentation of human motion into plausible motion primitives is central to the understanding and building computational models of human motion. Several issues contribute to the challenge of temporal segmentation of human motion. These include the large variability in the temporal scale and periodicity of human actions, the complexity of representing articulated motion, and the exponential nature of all possible movement combinations. We formulate the temporal segmentation problem as an extension of standard kernel k-means clustering, and derive an unsupervised hierarchical bottom-up framework called Hierarchical Aligned Cluster Analysis (HACA). HACA extends standard kernel kmeans clustering in three ways: (1) allows the cluster means contain a variable number of features, (2) introduces a generalized dynamic time warping (DTW) kernel as temporal metric between sequences, and (3) incorporates parameters that enable the user to specify the time granularity of the motion primitive in the hierarchical decomposition. Experimental results on motion capture data and video demonstrate the effectiveness of HACA for decomposing complex human motions.