Feature-based face recognition system using utilized artificial neural network

Abstract

This project aims to reduce the effect of critical conditions such as excessive illumination, facial expressions, hairstyles, beard and moustache which have affected the performance of face recognition since ages ago. The main contributions of this project are the automatic algorithms for mouth detection, facial features cropping and face classification. First, the algorithm will detect a human face and irises. Second, the mouth region is estimated by using geometric calculation based on the irises positions. A proposed algorithm which combines RGB color map and corner detection techniques will detect the mouth corners. Then, the proposed features cropping system will crop the detected iris and mouth automatically. These features are fed into the backpropagation neural network. The proposed architecture contains two neural networks. The second network merges the results from template matching and first neural network to reduce wrong recognition rate and improve the performance of neural network. The proposed automatic feature-based face recognition system has efficiency more than 95% under the stated critical conditions. All the experiment results are studied to prove the quality and uniqueness of this research.