Research On Face Recognition Based On Classifer Fusion

Face recognition technology has been attached great importance the theory research and practical value. In the past few years, face recognition become the hotspot of artificial intelligence and pattern recognition. It can be applied to human identification, digital monitoring, teleconference and so on. Face recognition be regard as three process, including face detection, feature extraction and pattern recognition. In this paper by analyzing current relevant algorithms, focusing on research two problems: features extraction and classifiers fusion. The main contributions of work as follows:(1) Fusing two different methods about feature extraction is proposed. One feature extraction method is using DDCT (Divided Discrete Cosine Transform) plus LDA (Linear Discriminant Analysis) extract part-area features, another feature extraction method is using PCA (Principal Component Analysis) plus LDA extract global features. At least, fusing two different methods based on the rule of fusion. It can help classify classes, because of fusing features adding more identification information for classifier.(2) The method of combining minimum distance classifier and support vector machine classifier is proposed based on the thinking of threshold determination. The method including two parts. Firstly, checking the human face image is the part of human face database or not. Secondly, if the face image is the part of human face database, calculating the degree of classification, and then depending on the degree of classification choose the minimum distance classifier or support vector machine classifier.