Texture-Based Face Recognition Using Grasshopper Optimization Algorithm and Deep Convolutional Neural Network

Abstract

Face recognition is an active research area in biometric authentication, which has gained more attention among researchers due to the availability of feasible technologies, including mobile solutions. However, the human facial images are high dimensional, so the dimensionality reduction methods are often adapted for face recognition. However, the facial images are corrupted by the noise and hard to label in the data collection phase. In this study, a new GOA-DCNN model is proposed for face recognition to address those issues. Initially, the face images are collected from two online datasets FEI face and ORL. Next, modified local binary pattern (MLBP) and speeded up robust features (SURF) are used to extract the feature vectors from the collected facial images. The extracted feature values are optimized using grasshopper optimization algorithm (GOA) to decrease the dimensionality of data or to select the optimal feature vectors. At last, deep convolutional neural network (DCNN) was applied to classify the person’s facial image. The experimental result proves that the proposed model improved recognition accuracy up to 1.78–8.90% compared to the earlier research works such as improved kernel linear discriminant analysis and probabilistic neural networks (IKLDA + PNN) and convolutional neural network (CNN) with pre-trained VGG-Face.