A Hidden Markov Model-based approach for recognizing swimmer's behaviors in swimming pool

This paper employs a HMM (Hidden Markov Model) methodology to recognize some specific-behaviors of a swimmer in the swimming pool. In this study, a swimmer's behavior is composed of a series of the static image-frames. For each frame, a Convexity-Structure is used to enclose the swimmer's shape after segmenting the swimmer-blob. A codebook is created for mapping a swimmer-blob into one of 6 feature blob-types during training and recognition stage. Consequently, the time-sequential blobs are converted to a feature-vector sequence and transformed into a symbolic-sequence by the existed codebook. Thus a learned HMM can be obtained by this symbolic-sequence from a given specific-behavior and used to recognize a swimmer's behavior which is normal or abnormal. Four different swimming styles (Backstroke, Breaststroke, Freestyle, and Butterfly) are examined. An average recognition rate of 90% is obtained.