A Novel Wavelet-Based Multi-Task Convolutional Neural Network and Genetic Algorithm Approach for Enhanced Facial Recognition

Abstract

A new system based on wavelet transform, a multi task convolutional neural network and feature extraction technology, as well as a genetic algorithm model, is put forward to address the insufficient accuracy and high data noise in current Facial Recognition (FR). The new system uses a wavelet transform based face detection algorithm to denoise the noisy data of facial images. Afterwards, feature extraction and genetic algorithm are used to improve the model and enhance the accuracy of facial image recognition. The study results demonstrated that the approach used in the study was 16.0%, 10.0%, and 8.7% higher than other algorithm models in Dataset 1, and 15.3%, 10.1%, and 13.3% higher than other algorithm models in Dataset 2. The highest change value in dataset 1 was 92.3%, which was about 1.1% and 4.7% higher than other models. The recognition accuracy, recall, and regression values of the research method model were the highest at 99.7%, 99.5%, and 99.7%, respectively. The accuracy of the model was 6.00% and 3.33% higher than that of other models. The stability of the research method was the best among traditional algorithms. The approach architecture used in the analysis has improved performance and higher accuracy in FR.