Dynamic gesture track recognition based on HMM

The dynamic gesture track training based on HMM (hidden Markov model) is one of the key techniques in dynamic gesture recognition. This paper adapts the iteration algorithm of Baum-Welch on the HMM to train and do some research to the performance of dynamic gesture track training from iteration times, sample number selection and model initial value selection. The experimental results show that the HMM is very efficient to the dynamic gesture track recognition with spatio-temporal characteristic.