An evolutionary single Gabor kernel based filter approach to face recognition

For the past few decades, Gabor filter banks have been used for extracting features in face recognition. The state-of-the-art approach for the design and selection of Gabor filter banks is based on a trial and error. This results in more computational complexity and higher response time. To overcome this problem, an attempt is made to design a single optimized Gabor filter instead of a filter bank for feature extraction. This approach improves the filter performance by significantly reducing the computational complexity and response time. The hybrid particle swarm optimization-gravitational search algorithm (PSO-GSA) is used for optimizing the parameters of the single Gabor filter. In this context, an evolutionary single Gabor kernel (ESGK) based filter approach is proposed for face recognition. The proposed method is used to extract Gabor energy feature vectors from face images. We also propose a new eigenvalue based classification approach for face recognition. This approach is derived from sparse based representation methods. The novelty in our paper is measurement of sparsity of the weighting coefficients of each training sample. The main contribution of the paper is two-fold. Firstly, investigation of ESGK approach, which is not found in the literature. Secondly, introducing a new eigen value based classifier. FERET, ORL, UMIST, GT and LFW databases are used to measure the efficiency of our proposed method. The results are compared with a holistic Gabor filter bank based recognition methods. It is witnessed that our proposed method outperforms the state-of-the-art methods.