Improving BP Neural Network for the Recognition of Face Direction

Abstract

the recognition of face direction is an important part of the artificial intelligence. In recent years, BP network has been used for pattern recognition. However, in practical application, BP has some disadvantage. The widely used BP algorithm has slow convergent speed and learning efficiency, and it is easy to get into local minimum. Selection of the initial value of the BP network can also affect convergent speed. This paper presents an improving BP network to accelerate convergence with genetic-simulated annealing algorithm. So, we optimized the initial value of the network through adding annealing idea into genetic algorithm(Genetic-stimulated annealing algorithm, GSA) to identify face direction. Using this improving BP neural network for the recognition of face direction, the results presents that our method has the highest precision and reaches relatively good effects compared with traditionally BP neural network.. Therefore, this method optimized with GSA poses better recognition ability, and achieves ideal effect for the face direction.