Effect of various distance classifiers on the performance of bat and CS-based face recognition system

Abstract

Due to increasing risks mainly in surveillance, security, and authentication, it has become imperative to pay more attention to Face Recognition (FR) system. Any FR system consists of three main subdivisions—Feature Extraction, Feature Selection, and Classification. This paper uses a combination of DCT (Discrete Cosine Transform) and PCA (Principal Component Analysis), i.e., DCTPCA for feature extraction followed by Bat and Cuckoo Search Algorithms for Feature Selection. The aim here is to use different classifiers such as Euclidean Distance (ED), Manhattan Distance (MD), Canberra Distance (CD), and Chebyshev Distance (ChD) for classification purpose and to compare them to find as to which, among these, suits best for a given dataset. The results not only reveal the efficiency of Bat-based feature selection algorithm over Cuckoo Search but also show how effective Euclidean Distance classifier is over other classifiers for Yale_Original database and Manhattan Distance classifier for Yale_Extended database.