Action Recognition in Motion Capture Data Using a Bag of Postures Approach

In this paper we introduce a novel method for movement recognition in motion capture data. A movement is regarded as a combination of basic movement patterns, the so-called dynemes. Initially a K-means variant that takes into account the periodic nature of angular data is applied on training data to discover the most discriminative dynemes. Each frame is then assigned to one of these dynemes and a histogram that describes the frequency of occurrence of these dynemes for each movement is constructed. SVM classification and sparse representation based classification are used for movement recognition on the test data. The effectiveness and robustness of this method is shown through experimental results on a standard dataset of motion capture data.