Speaker-identification Algorithm Research Based On Lip-features

Speaker-identification technology based on lip-features has achieved rapid progress in modern society, especially in the information systems used in government, military, banking and other agencies. Speaker-identification technology based on lip-features attracts widespread attention because of excellent real-time performance, ease of collecting samples, etc.In order to overcome limitations of current lip-feature extraction algorithm, including heavy computation in extracting lip-features and low rate of recognition, we have proposed a new speaker-identification algorithm based on lip-features, which has been thoroughly tested and examined in the lab. The primary tasks carried out are listed as follows:1.On analysis and research of existing identification based on bio-features and lip-features, we conceived lip-feature data extraction and identification by using interpolation, and designed the architecture and design of the related components.2. In extracting of lip-feature data, the Paper adopted a strategy of geometrically integrating static and dynamic lip-features, proposed lip-feature data extraction by using interpolation, and tested all major aspects of the proposed algorithm.3. In identifying the speaker, we designed an identification method based on the lip-feature data. Through screening the analogical sets by way of setting up relevant thresholds, and then differentiating the similarities, the speaker’s ID can be locked.4. Availability and effectiveness of lip-feature data exploitation based on interpolation was verified through simulation experiments. Identification of the speaker by way of lip-feature data exploitation showed a high rate of recognition, thus proving the scheme’s excellent effectiveness.