The Design Of Lip Feature-based Identity Recognition Algorithm

Lip recognition is a lip movement feature-based retrieval technique of the speaker’s information. It has a high application value in the fields of speech sound recognition, identity recognition, intelligent human-machine interface as well as multimedia system etc.. Since most existing lip-feature extraction are based on theoretical research of extraction algorithms and Little discussed from the application system design, according to this case, the present paper introduces in detail the design of a lip feature-based identity recognition system. the main work of which includes:(1) The overall design of lip feature-based speaker’s identity recognition system, which consists of video information collection and preprocessing, lip feature information extraction, identity recognition and auto-learning, is made after an analysis of relative skills as well as general principles.(2) An integrative representation method of both geometrical and dynamic lip features is applied here in the process of lip feature information extraction. The algorithms in modules of feature information extraction process, relevant feature points selection, side selection and length calculation, motion vector extraction as well as key points trajectory fitting are designed first.(3) Lip feature-based speaker identity recognition process is then designed and discussions are also made on algorithm design of speaker’s lip feature reconstruction, sample sets selection, movement similarity and difference analysis.