Robust Swimming Style Classification from Color Video

Abstract

We present a robust method to classify swimming styles from live video based on the features extracted from the upper-body of the swimmer. In our approach, potential body parts are first extracted using a simple skin color model. The segmented regions are further analyzed to isolate the arm and shoulder blocks. Finally, a scoring system based on quantitative measures such as the slope, size, aspect ratio and the relative position of the body parts is constructed to carry out the classification. Several enhancements to the original scoring system are developed and tested. Experimental results demonstrate the validity and efficiency of our proposed approach.