Speaker Recognition System Research Based On Double VQ-SVM

Speaker recognition is a biometrics that the identifier of a person can be recognized via his voice. It is applied to man-machine interface, ensure public security, military affairs, judicature, and so on. Speech signals preprocessing, speech enhancement and feature extraction were discussed. A two-level-architecture recognition method combining double vector quantization (VQ) and support vector machines (SVM) was proposed, which could improve the recognition rate. A threshold was adopted to improve the performance of the system. A variation step was adopted to modify the threshold and improve the efficiency of the system. The result of the experiment shows that compared with double VQ recognition method the proposed method can highly improve the recognition rate under clean speech and noise environment. The system based on double VQ-SVM is useful and effective in speaker recognition.