In communication and the processing of signal, it is inaccurate that noise can be seen as Gaussian noise and multi-model noise. Gaussian Mixture Models is an efficient method of simulating complicated density with simple structures and present a natural frame and semi-parameter structure of modeling unobserved population homogeneity and heterogeneity, and multi-model noise is a special example. Gaussian Mixture Models has good clustering performance. EM algorithm is one of many parameter estimation algorithms of Gaussian Mixture Models. It is Maximum likelihood estimation based on incomplete data. EM algorithm highly depends on initial parameters, it can not estimate the number of components of models, and the covriance matrix get into singularity. The higher-order statistic of Gaussian Mixture models is efficient algorithms, and helpful to the parameters initialization of Gaussian Mixture models estimation, especially the skewness and kurtosis. Annealing algorithms has local searching ability, and it is applied to EM algorithms to be hybrid EM algorithm. Genetic algorithm has strong global searching ability, so the algorithms of combination of GA and hybrid EM algorithm can estimate the numbers of components of models and parameters of GMM efficiently. Then the algorithm is applied to clustering with UCI data, and the accuracy rate is compared. |