Recognition of Human Motion through a Fuzzy Hidden Markov Model

A new type of Hidden Markov Model (HMM) developed based on the fuzzy clustering result is proposed for identification of human motion. By associating the human continuous movements with a series of human motion primitives, the complex human motion could be analysed as the same process as recognizing a word by alphabet. However, because the human movements can be multipaths and inherently stochastic, it is indisputable that a more sophisticated framework must be applied to reveal the statistic relationships among the different human motion primitives. Hence, based on the human motion recognition results derived from the fuzzy clustering function, HMM is modified by changing the formulation of the emission and transition matrices to analyse the human wrist motion. According to the experimental results, the complex human wrist motion sequence can be identified by the novel HMM holistically and efficiently.