Study Of Dynamic Face Recognition System On Wavelet Transform

Along with the development,of informative process,digitalization and network,the safety of national as well as social activities is faced with new challenge.Under this kind of environment,traditional safe technology shows a lot of abuses and defects.Biological feather certification technology is the one of better method of solving the safe problem of this new challenge.Face recognition has its specific advantages,may be extensively applied in the department of public security,convict identify search,punishment detective case,safe verification and supervisory system,the human-computer interaction control, certificate check,passageway control and verification of credit card.The paper is mainly concerned of the preliminary study of the dynamic face recognition from the face image preprocessing,the establishment of multi-profile face characteric database,extract facial features by wavelet transform and classification methods to identify on the non-restriction environment.The face image preprocessing is a very important step of face recognition on the non-restriction environment.The paper is introduced a detailed account of the facial image of the geometric correction,gray processing,light processing and shelter processing,and facial image preprocessing to improve recognition rates role by comparing the experimental verification.The paper is introduced a detailed discussion of facial features arithmetic,Spectrofaces arithmetic and wavelet transform coefficients arithmetic on extract facial features by wavelet transform, and analysis the effect of the different layers and radixes of wavelet transform to face recognition rate and speed.We know the biorthogonal wavelet transform has the best result on face recognition.The paper is introduced a biorthogonal wavelet,the use of pretreatment of the trainning images to establish a human facial characteric database,and has a ideal face recognition result by contrast the other types of wavelet transform.In the paper,we introduce the principle expounded of support vector machines and Euclidean distance classifier,while improve Euclidean distance classifier on the basis.Facial characteric database is the source of data of face recognition.The paper is introduced the establishment method of a facial characteric database by using five different facial gestures.For recognition image is matching with the nearest pose,and it can deal with dynamic face recognition under the small angle case.The paper uses the different geometrical characteristic classification of the eyes,mouth,nose and face figures to divide a face database into many sub-database by processing,and it can minish recognition time.In the emulator of face recognition,we establish a facial characteric database increasing face recognition rate by improved Euclidean distance classifier.We can be seen from the emulational results that face images pretreatment has a important role to face recognition rate,the use of five different facial gestures to establish a facial characteric database is a good method to solve the problem of dynamic face recognition,wavelet transform can effectively improve face recognition rate and speed, different wavelet transform layers and radixes on face recognition results have a important role to face recognition rate and speed,and improved Euclidean distance classifier has a Unique advantage in many swatches compared with other classifications.