Research On Speaker Feature Modeling And Application In Information Security

With the fast development of information technology, the personality identification based on safe access control is more and more significatively in praxis. The speaker recognition is one of the kinds of biometrics technology, which has caught much attention for its particularly advantage on convenience, economy and veracity. It become an important and popular authentication technique in human life and work consequentially. However, the modeling of speech signal characters and the research of automatic speaker modeling is still far from practicality. Therefore, a more robust method for speaker recognition with high accuracy of recognition rate is imperative under the situation.Among numerous speaker recognition approach, this paper focuses on the speaker recognition system based on common methods of feature extraction, Dynamic Time Warping and Hidden Markov Model, and then a approach combined with the Clustering Analyse and Principal Component Analysis is proposed. This approach reduces the correlation of high dimension parameters, enhance the speed of disposal. Meanwhile, it makes the partition of the range about the characters more veracious by weight. Experimental results show that the new approach is much more close to the human proper characteristic. It can improve the accuracy of recognition rate effectively and good practicality value.