Face Recognition System Besed On DSP

Face recognition is an attractive research subject that has very important theoretical value and practical value in the pattern recognition fields; it has wide application in the human ID management, intercourse between human and machine, video surveillance and so on.This subject study the technology of face tracking and recognition to realize the independent system based on DSP/BIOS. This subject has done the research in many aspects such as image preprocessing, face tracking, feature extraction face tracking and detection.In the stage of Face tracking and detection, the face is detected by the use of complexion information. After set the image as two value and eliminate noise the article locate the center and the face area. After line and row projection .both position of eyes is located. Set the distance between two eyes as standard distance, then cut the human face according the geometry person make the use of level accumulate points and perpendicularity accumulate points. Take the distance between both eyes as the standard distance, according to the geometry characteristic .set the face image dimension normalization and get the standard face image..In the course of character extraction, had contrasted the DCT shifts with the wavelet shifts. Adopt the lifting scheme wavelet transform, after three layers wavelet transform get the low frequency information as face feature.the lifting scheme adopt the same address for wavelet transform has cut down the needing space and realized the system of face recognition.The system adopt DSP/BIOS real time OS, manage all kinds of resource in the whole system.The result show that the method brought forward by this article has some agility under some comparatively complicated environment. The method can track human face fairly well, the recognition result can achieve 75% when detect the eyes correctely. The result is demonstrated that the method has certain pragmatism.