Optimization of HMM parameters based on chaos and genetic algorithm for hand gesture recognition

In order to prevent standard genetic algorithm (SGA) from being premature, chaos is introduced into GA, thus forming chaotic anneal genetic algorithm (CAGA). Chaos' ergodicity is used to initialize the population, and chaotic anneal mutation operator is used as the substitute for the mutation operator in SGA. CAGA is a unified framework of the existing chaotic mutation methods. To validate the proposed algorithm, three algorithms, i. e. Baum-Welch, SGA and CAGA, are compared on training hidden Markov model (HMM) to recognize the hand gestures. Experiments on twenty-six alphabetical gestures show the CAGA's validity.