Face Recognition Based on Weighted Fusion of Face Similarity Features

A method for weighted fusion of similarity features to describe faces is proposed. Similarity feature is extracted from similar face, which has good stability. The fusion of uncorrelated similarity features is more conducive to face recognition. Firstly, the Euclidean distance between similarity features is calculated, and the similarity features are screened to find a set of uncorrelated similarity features, which effectively reduces the number of features and removes redundant information between similarity features. Then, the similarity features are adaptive weighted fusion and the fusion result is detected in the AR face database. The final simulation experiment achieve good results.