Research Of Speaker Identification Based On The Lip Movement

Biometric identification has rapid development in defense, military security and civilian authentication, especially identification based on the dynamic lip receives widespread attention due to its unique advantages.The tradition feature extraction of lips mainly based on single static image, which did not full use of the feature of lip movement, and neglected the problems of huge information and low recognition rate. For these deficiencies, we designed a method of speaker identification based on critical phonemes. The main work and researches are as follows.(1) On the basis of the general principles and technologies of identification, a scheme of speaker identification based on critical phonemes is constructed, which composed by three parts:video preprocessing, lip feature information extraction and identification.(2) In the lip feature information extraction stage, the paper design the pattern of speaker’s characteristic which integrates the speaker’s lip static and dynamic feature and construct a model of lip feature information extraction. The functional relationships of each part of model and specific processing algorithms are given.(3) In the speaker identification stage, we design and construct a universal phoneme model, then design a feature phoneme model of the speaker, finally a complete identification model based on critical phonemes are given. We detailed design algorithm including universal phoneme extraction, feature phoneme extraction and speaker identification.(4) Simulation experiments verify feasibility and effectiveness of lip feature extraction algorithm, and analyze the performance of speaker identification algorithm. The experiment results show that identification program designed in this paper can better solve the problem of large feature information and high redundancy rate in identification.