Implementation real-time gender recognition based on facial features using a hybrid neural network Imperialist Competitive Algorithm

摘要

In this paper, we have proposed a robust approach for developing an automatic system for gender classifying from a facial image at video files or image files. In the proposed algorithm, feed-forward artificial neural network manages the classification problem by use of Imperialism Competitive Algorithm (ICA) to achieve the best weights of the neural network instead of old gradient based methods, like back propagation. In this paper we first extract some facial segments from given videos by Viola Jones algorithm. Then feature selection through Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) is done and finally a hybrid Artificial Neural Network (ANN) with ICA (ANN-ICA) performs classification. Experimental results show that combining the feature extraction techniques with the ANN-ICA for classification, the performance of gender classification improves significantly and reached a recognition rate of 94.3%.