A Novel Fuzzy Kernel Vector Quantization for Speaker Recognition with Short Utterances

When the amount of available training and testing data will be few seconds, the number of feature vectors we obtain are less which are insufficient to model and discriminate speaker well. It presented a new method for speaker recognition with short utterances. By non-linear mapping, it used the sectional set fuzzy Vector Quantization with Lp norm to form speaker´s model in the high-dimensional feature space. During the recognition phase, it used the sectional set fuzzy kernel nearest prototype classifier to identify unknown speech. Kernel mapping made the inherent speech features explored, and the dissimilarity among different speakers increased. It used Lp norm to replace the Euclid norm in Gaussian kernel function, which make the algorithm more robust. In order to improve the convergence speed and recognition rate, it also used the sectional set method to modify the membership function. Experimental results show that for about 5s of training and 1s of testing data, the performance of proposed method are 94.83%.