The System Of Based On Sparse Coding Face Recognition In Video

With the development of modern science and technology, video data growsexplosively. These huge amounts of video data, especially the mass person identity fromvideo data has been of great value, which would require face recognition technology invideo; However, current research on face recognition in video is still far from enough.The research of face recognition in video can be used for secure area and controlsystems, particularly with the advent of mobile intelligence era, mobile terminals caneasily capture large amounts of video data, which contains a large number of peopleinformation, which requires of efficient face recognition in video, combined withpowerful computing power of mobile terminals on these information digging andanalysis. Therefore, the research of face recognition in video is very meaningful.Face recognition in video is to classify the faces in video, its main researchchallenge is that the face data is easily effected by the facial expression, illumination,resolution of video, occlusion, such worse, resulting in some of the traditional faceclassification algorithm is not effective in video. The research subject of this paper is theface recognition in video based sparse coding, mainly focused on effective facerecognition algorithm for video. Analysis the characteristic of human faces in video,sparse coding ideas introduced in the video face classification. This paper uses theLC-KSVD, which is a face classification algorithm based on sparse coding, classify theface for video. Meanwhile, fully consider video facial characteristic, optimize theoriginal LC-KSVD algorithm to improve the classification performance in video. Themain methods are: based on video sequence, building the dictionary element for sparsecoding; modifying sparse coding consistency matrix; statistical classification used whenvoting policies, improving accuracy. And comprise the optimized LC-KSVD algorithmand the tradition face classification algorithm for video, and find it is more effective.Finally, design and implement a video face recognition system, including shotsplitting, face detection, face tracking, speaker tagging, face feature extracting andclassification. The paper makes some innovation for some steps of the system, mainlylying in doing second detection during face tracking step, to make the face tracks longer,and making a filter for face when face tracks extraction.