Speaker recognition based on Bayesian network
Hongjie Wan · Jisuanji gongcheng · 2005
Based on Bayesian network theory and according to the task of speaker recognition with text independent, a Bayesian network structure for speaker recognition was presented. Methods of constructing the network,parameter estimation and how to inference and classify in speaker recognition were also given in detail.Experimental results under the same train and test data show that this network structure can achieve thirty percent average relative improvement in performance than classical Gaussian mixture model.