Abstract: In view of the shortcomings of high error recognition rate and poor robustness of current face recognition algorithms, we designed a face recognition algorithm based on discriminant analysis and low rank projection to obtain better face recognition results. Firstly, the face images were divided into blocks, the local feature of each sub block was extracted, and the global feature of face was extracted by discriminant analysis. Secondly, we used low rank projection to make up a feature vector that contributed greatly to the face recognition results. Finally, the least squares support vector machine was used to establish multi classifier for light face recognition based on the principle of onetomany, and the simulation experiment was carried out by multiple face database. The experimental results show that the algorithm can find the optimal feature subset of face recognition, reduce the error recognition rate of face, the speed of face recognition is obviously improved, and the effect of face  recognition is better than other face recognition algorithms.

Key words: face classifier, discriminant analysis, speed of recognition, face recognition, feature selection