Speaker Verification Based on TES-PCA Classifier and SVM plus FCM Clustering

Speaker verification is an important branch of speaker recognition. In this paper, a novel hierarchical speaker verification method based on TES-PCA Classifier and support vector machine plus Fuzzy c-means clustering was proposed for the sake of improving performance of speaker verification. In this algorithm, we utilized PCA and Fuzzy c-means clustering to select more discriminant and lower dimensional feature vectors firstly. And then, the truncation error space(TES) was obtained from PCA transformation matrix. The R target speakers were selected fleetly from TES-PCA classifier. Finally, support vector machine was used to make final decision. The experimental results showed that our proposed method could improve recognition accuracy remarkably and the system has better robustness compared with traditional methods.