Research Of The Speaker’s Identity Verification Technology Based On Voiceprint Recognition

Identity verification is one of the most important parts of information security. The traditional ways of identity verification is in the form of password, which exposes a lot of drawbacks. The technology of using one’s biological feature to verify identity develops gradually and has become the new research hot point. Voiceprint recognition as one of the most important branch of biological recognition, has been paid close attention to, because of its splendid characteristic such as, convenient, inexpensive and precise, etc. Now, it has been used in the fields of e-commerce, administration of justice, encrypt passwords, etc.Voiceprint recognition, is the technology of automatic verification of speaker’s identity according to the feature obtained from one’s voice. It focuses on recognizing speaker, rather than what the speaker says. This paper researches identity verification technology based on the voiceprint recognition. And its purpose is pushing the technology to become practical. At first, the basic knowledge of voiceprint recognition has been introduced. Then, I study three important parts of voiceprint recognition system: pretreatment, feature extraction and pattern matching. The wavelet transform theory is used to improve traditional MFCC feature extraction, and VQ and HMM are combined to create a new method of pattern matching. And all the methods above are used to build a identity verification system. Finally the analysis of experimental results are provided.Summary, the voiceprint recognition system works very well. And identify speed is fast, the cost is cheap. But it also has some shortage: the model’s training time is too long. After all, I hope that after some deeper research, the recognition system can be improved. Voiceprint recognition technology has wide market prospect, and our county is devoting more effort to it, so this technology will be riper and more perfect in the future.