An evolutionary learning model based on FSS and eigenspace approaches for distance-alterable face recognition

In pattern recognition, the performance is almost limited by the quality of data and influence of environment. And, it is hard to consider all variations of data in advance. Thus, in this paper, we propose an evolutionary learning model which can utilize basic information to absorb related information. The performance can be improved because knowledge gets richer. However, if the wrong information is learnt, the performance will decay seriously. In order to avoid the problem, a tutor scheme is necessary. Therefore, we propose a Feature Selection Strategy (FSS), which can sieve out the redundant and wrong information. Besides, an eigenspace approach called ILDA and corresponding distance measure method are adopted to assist FSS and recognition. In the experiments, we apply the evolutionary learning model in face recognition system. Few clear and straight face images are provided as basic information of subjects, and face images shot in different distance are recognized with learning model or not. By comparing the accuracy, we verify the performance of the proposed evolutionary learning model is suitable for distance-alterable face recognition. Otherwise, we replace ILDA with PCA, and repeat the experiment again. The experimental results indicate the feasibility of the evolutionary learning model.