噪声环境下文本无关说话人识别研究

his paper researches of improved spectral subtraction method of speech enhancement algorithm by mending the question that the recognition rate is low in text-independent speaker recognition under noisy environment. This algorithm to be applied in front-end process of speaker recognition system, to achieve noise reduction for input speech signal. Analyzes and experiments speaker recognition rate with methods of feature extraction of MFCC and recognition of GMM. Experimental results show that the performance of speaker recognition rate is improved with the processed noise reduction for speech signal.