This article puts forward a model based on decision-level fusion of multimodal biometrics. Combination of principal components analysis (PCA) and linear discriminant analysis (LDA) to identify the face images of people. The combination algorithm of Mel frequency cepstrum coefficient (MFCC) and Gaussian mixed model (GMM) were realized to the speech recognition. The results of the face recognition sub-system and the speech recognition sub-system were chosen as the inputs of support vector machine (SVM) in decision-level. Fused by a linear kernel SVM classifier

finally results would be obtained. It shows many good properties in classification

improving the generalization ability

reducing the number of support vectors. It can effectively enhance the rates of the whole system.