Speaker recognition is a process of identifying the speaker automatically based on parameter which reflects the physiological and behavioral characteristics in speech waveforms. As a kind of biological technique, speaker recognition applied to Internet Communications, bank and so on. Speech signals preprocessing, feature extraction and methods of anti-noise were detailed discussed in the paper. The methods of speaker recognition based on principal component analysis (PCA) and least squares support vector machine (LS-SVM) were detailed researched. For large crowd speaker recognition system, the speaker recognition model based on P&T+LS-SVM was constructed and the algorithm process was detailed introduced. For performance of the speaker recognition model, a lot of experiments were done. The result of the experiment shows that the speaker recognition model can achieve good performance and has good robustness in noisy environment. |