Harris Hawks Optimization Method based on Convolutional Neural Network for Face Recognition Systems

This paper discusses the momentum gradient dependent on the convolutional neural organization's strong point. It is a new methodology introduced to detect evenness in the data set of faces. The proposed face recognition framework was created for various purposes. Through Gabor wavelet change, facial evenness was extracted from the face-preparing information. After that, we applied a profound learning process to carry out verification. After applying the proposed method to YALE and ORL data sets, we simulated them using MATLAB 2021a. Before this, similar trials were directly applied through Harris Hawks Optimization (HHO) for including the determination approach. The extraction process was conducted with many picture tests to execute the Gabor wavelet method, which proved more viable than other strategies applied in our examination. When we applied the HHO on the ORL dataset, the acknowledgment rate was 93.63%. It was 94.26% when the three techniques were applied to the YALE dataset. It shows that the HHO calculation improved the exactness rate to 96.44% in the case of the YALE dataset and 95.88% in the ORL dataset.

I. Introduction

Face processing is a main problem in machine vision algorithms, which uses intelligent methods to identify human beings based on their physiological characteristics. This technology has advantages such as high accuracy and low level of individual intervention. It has benefits in many applications regarding information security, law enforcement, monitoring, traffic control, and registration in attendance systems. In recent years, in-depth learning has achieved tremendous performance in computer vision, especially in terms of face recognition. Deep learning uses a combination of abstract features and deep neural networks to generate results that are generally unachievable and indistinguishable. The structure of the deep network-based face recognition model depends on the training data quality and number. In general, face recognition systems have challenges such as changes in brightness, different facial expressions, face coverage, low image resolution, and limited examples [1] [2].