Design and application of face recognition algorithm based on improved backpropagation neural network

Abstract

Face recognition is a promising technology with a great application potential and broad prospects for development. Compared with other identification technologies, face recognition can achieve rapid and easy sampling, without affecting the behavior of the sampled. These advantages have induced a surging demand and interests in this technology, making it a research hotspot in artificial intelligence. This paper extracts the features from the target face image by Principal Component Analysis (PCA), reducing the dimension of the image. Taking the feature coordinates of the face image for classification, it is possible to eliminate the excess computing load induced by high dimensionality. After that, the backpropagation (BP) neural network was improved by the scaled conjugate gradient (SCG) algorithm. The improvement aims to control the model error caused by the defects of the original BP neural network, including inefficient learning, slow convergence and proneness to local minimum. The improved BP neural network was then adopted to classify the feature coordinates of the face image. Finally, the proposed face recognition algorithm was implemented on Matlab and trained with the improved BP neural network. The experimental results show that the proposed algorithm achieved good recognition performance.