Text-Independent Speaker Recognition System

• Main Authors: Mao-Chieh Tsao, 曹懋杰
• Other Authors: KUO,JIEH-SEN
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/11403323981516774027

碩士 === 高苑科技大學 === 電子工程研究所 === 100 === Organism recognition technology utilizes the human characteristics to identify the operator, as well as the speaker recognition system performs the identification via using the voice feature of speakers. Recently, speaker recognition technology was applied to several industries, such as the entrance guard system of building, the net work security, and some special situations which need to recognize the people. The human voice is used to establish the recognition system. Get along with the improvement of voice recognition technology, text-independent speaker recognition technology raises more and more related research and get more and more attention on it. Thep urpose of this thesis is to establish a text-independent speaker recognition system with few sparkers. There was 30 voice samples used in this research. Each sample recorded the voice 60 seconds in Chinese. After preparing the samples, we first extract the feature of average spectrum, and then utilize the Euclidean distance and Mahalanobis distance theory to be the identification device, finally calculate the decision rule by using minimum distance theory. The results indicate that the Euclidean distance and Mahalanobis distance techniques in text-independent speaker recognition system have much efficiency to recognize the speaker accurately.