Face Recognition: Innovative Advances and Future Impact

Abstract

Images in which secondary channels are obtained in several levels or spectral extents of the Electromagnetic Scale (EMS) are referred to as multidimensional images (also known as hyperspectral or multispectral). Everyone has a claim to information that can be manipulated in face recognition (FR) applications. Apart from Visible Spectrum (VS) data, a facial investigation system can also utilize infrared (IR) imagery to identify presentation attacks, stance variations, and facial look dissimilarities. This research presents an evaluation of current multi-modal face recognition (MDFR) algorithms using imagery from VS, the Near Infrared (NIR), Short Wavelength Infrared (SWIR), and Long Wave Infrared (LWIR) sub-bands of EMS. An assessment about their requirement is conducted, and the public multispectral and hyperspectral datasets for facial analysis are acknowledged. The basic operating standards of the MDFR approaches are categorized, ranging from the traditional Fusion and Subspace schemes to the most recent Deep Neural Networks (DNN).