Robust face recognition from single training image per person via auto-associative memory neural network

Face recognition from single training image per person is one of important challenges in appearance-based pattern recognition field. Although many existing face recognition methods have achieved success in real application, but can not be directly used to the single training image scenario. The associative memory neural networks provide a feasible strategy to address such problem. In this paper, we first briefly review the existing single training sample face recognition algorithms, and then propose a new multiple value auto-associative memory neural network by modifying evolution rule and activation function. Finally, experiments on the two publicly available face databases are provided to validate the feasibility and effectiveness of the proposed algorithm.