<p>Due to the complex nature of facial images and the presence of noise, face recognition is still a hard task in computer vision. In order to solve these issues, this paper suggests a new system that uses specific feature extraction along with deep learning and metaheuristic optimization. To accomplish this, facial features are taken using Gabor Wavelet Transform (GWT) together with Haralick texture descriptors, both capturing frequency and texture. The system makes use of the Honey Badger Firefly (HBFF) algorithm to choose the best and most important features, which lowers the size of the data and improves the recognition rate. A Deep Learning Neural Network (DLNN) is used to identify the chosen features and group them for face recognition with good accuracy. Through utilizing the crucial datasets like ORL and YALE reported that the proposed method performs better in terms of accuracy, sensitivity, and specificity when compared to existing CNN–based approaches. Such results demonstrate how this method could be effective and dependable for facial recognition in everyday use.</p>