Swimming Style Classification from Video Sequences

In this paper, we present a novel method to classify swimming styles from video sequences based on the features extracted from arms and shoulders. In our approach, potential body parts are first extracted using a simple skin color model. Next, the segmented regions are filtered based on quantitative measures such as the aspect ratio and the area to retain the dominant parts. The resulting data is then down-sampled and regression analysis is performed to calculate the relative position of the constituting body parts. Finally, a decision tree is constructed to carry out the classification. Experimental results demonstrate the validity and efficiency of our proposed approach.