On this basis of summary the work of the people of past, this paper has introduced the basic knowledge of voiceprint recognition and basic principle in detail at first. Then the paper has mainly done the following works: according to the discrete time domain model produced by the pronunciation signal model, the pronunciation signal model will be carrying on pretreatment, extreme point measuring and feature extraction of voiceprint—LPCC, MFCC;
Summary, the voiceprint recognition system concerning text, whose identify rate attain to 83% when test six people, and identify speed is quick, and the speech sample collects easily, the cost is cheap.
Speaker Recognition (Voice Print Recognition) technology has been widely used in the information security domain due to its unique advantages. In this paper, the concept of speaker recognition, the design principle and the realization of GMM-based speaker recognition system are introduced. New feature extraction method for GMM-based system in a noise environment is presented. The tests show that the recognition rate is over 96%.