A Survey on Feature Engineering Based Human Face Recognition

Recognizing human faces has been a prominent focus of research for several decades. It has gained considerable interest and attention from various fields such as computer vision, artificial intelligence, and machine learning owing to its impressive advancements and wide-ranging societal applications. The primary goal of a face recognition system is to accurately identify human persons from still photos, video streams, and data streams to leverage contextual information associated with the active use of these data components. In this survey paper, we give enhanced commentary on local and holistic feature extraction or engineering methodologies or descriptors. To make a viewpoint on why these kinds of feature extraction are needed, we also discuss the challenges like illumination variations, pose variations, accumulation of noise, and occlusion that one has to face to build a robust human face recognition system. Also, we discuss standard face datasets that would possibly give those challenges to researchers to tackle to pursue human face recognition tasks. These datasets are proven handy in researching features for human facial textures. At the very end, we also try to start a discussion on why understanding better features and extracting them could be more beneficial to achieve higher accuracy in unconstrained environments rather than depending on just deep learning models. It is shown that enhancement in human face recognition can be achieved by extracting features and then seamlessly integrating the robust capabilities of deep learning models. This amalgamation is poised to yield superior results in accurately identifying individuals.