Fusing mixed visual features for human action recognition

Human action recognition has gained a lot of interest in the computer vision community in the last decades. A number of action classifiers have been developed for human action recognition, in which how to effectively represent high dimensional human actions for categorisation or recognition is a crucial factor. This paper presents a simple but efficient action recognition algorithm using mixed visual features. The mixed features fuse three action descriptors, namely centre distance-based Fourier descriptors, shape parameters-based regional descriptors and joints-based polar coordinates descriptors. The frame-based human action classifier is developed using random forests algorithm. Experimental results show that the proposed method is accurate, efficient and robust, and the combination of the three types of descriptors achieves superior performance in action recognition.