The Design Of Identity Recognition Algorithm Based On The Outer Feature Of Lip

With the rapid development of the Internet,the fast and effective verification of identity information becomes increasingly urgent which also attracts widespread attention in academic circles.Among them,the identification based on the biological features is remarkable because of its easy acquisition and not easy to be tampered.Aiming at the limitations of the feature parameters and incomplete characterization of existing lip feature extraction algorithm of biometrics,we designed and constructed an identification scheme based on the features of the outer lip,and researched and explored related aspects thoroughly.Now,we represent our main work in the following aspects.(1)Analyzing the research results of the existing lipreading recognition,and then designing a lip identification scheme at the basis of analysis which includes video preprocessing,outer feature extraction of the lip,outer feature measured human identity recognition process.(2)In the video preprocessing stage,we designed a detailed process of video preprocessing,which includes video segmentation processing,image processing,edge extraction and other aspects.In order to enhance the visibility of the outer lip,we designed a optimization method for edge contour of lip.(3)In the outer feature extraction of the lip stage,we designed the static and dynamic extraction models.Besides,we did some works aiming at the tracking of frames' key points,spline interpolation using key points,reconstruction function,topological construction for key points and other aspects.In order to improve the characterization for edge contour feature of lip and depict the change process of the pronunciation of the lips,we designed a new feature extraction method that contains static feature and dynamic feature.(4)In identity recognition stage,we designed a detailed process,which includes generated candidate set I and II,and the comparison of similarity.In order to improve the recognition accuracy,we designed a method which selected approaching values constantly by reducing samples.(5)In order to verify the feasibility and effectiveness of the scheme,we verify and compare it from the sieving rate of static features,the recognition rate of dynamic features and the overall scheme.The results of the experiment show that the scheme is feasible and effective.