Fractional Bat and Multi-Kernel-Based Spherical SVM for Low Resolution Face Recognition

Abstract

Face recognition is an important aspect of the biometric surveillance system. Generally, face recognition is a type of biometric system that can identify a specific individual by analyzing and comparing patterns in the facial image. Face recognition has distinct advantage over other biometrics is noncontact process. It has a wide variety of applications in both the law enforcement and nonlaw enforcement. While using the low resolution face images, the resolution of the image gets degraded. In this paper, to enhance the performance rate for low resolution image, the fractional Bat algorithm and multi-kernel-based spherical SVM classifier is proposed. Initially, the low resolution image is converted into the high resolution images by the kernel regression method. The GWTM process is utilized for the feature extraction by the Gabor filter, wavelet transform and local binary pattern (texture descriptors). Then, the super resolution images are applied to the feature level fusion by using the fractional Bat algorithm which comprises of fractional theory and Bat algorithm. Finally, the multi-kernel-based spherical SVM classifier is introduced for the recognition of feature images. The experimental results and performance analysis evaluated by the comparison metrics are FAR, FRR and Accuracy with existing systems. Thus, the outcome of our proposed system achieves the highest accuracy of 95% based on the training data samples, stopping criterion and number of draw attempts.