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

Abstract

: Face recognition is a unique feature for recognizing the individual in the biometric system and is advantageous since face recognition is a non-contact process. However, biometric recognition is ineffective due to the low-resolution images, wanting the need for the effective recognition system. Accordingly, this research concentrates on developing an effective face recognition strategy using low and high-resolution images. Initially, the input low-resolution images are pre-processed for enhancing the image contrast and subjected to the generation of the high-resolution image. Then, the feature extraction using the GWTM process presents the texture features that facilitate effective recognition using the spherical Support Vector Machine (SVM) that works using the multiple kernel function. In the GWTM process, proposed fractional-ROA is engaged in the optimal fusion of the features acquired from the wavelet, Linear Binary Patterns (LBP), and Gabor filter. The analysis of the recognition method is initiated based on the metrics, such as False Alarm Rate (FAR), False Rejection Ratio (FRR), and accuracy. The proposed fractional-ROA-based face recognition acquires the maximal accuracy, and minimal FRR and FAR of 0.98, 0.0123, and 0.0017, respectively.