Heuristic Optimization on Deep Neural Network with Horse Herd Optimization for Efficient Face Recognition Framework

Face recognition (FR) is the hotspot research area in image processing. FR technique reaches more attention and is widely used in various applications, where it emphasizes to identify the person by utilizing the face images. As it possesses more descriptive constraints, the implementation of FR is a big challenging issue. However, the most pervasive constraints are lighting effects, age variation, outliers and noise, different face angles and expression, which restricts to design an effective method. Earlier models are still facing the issue as extracting the relevant features of image; it is to be result with misleading the recognition process. To alleviate such issues, a novel deep learning model is proposed using heuristic algorithm. Initially, the image acquisition is carried out by two benchmark data sources. After collection, the preprocessing is achieved by median filtering technique. Consequently, the discrete wavelet transform (DWT) is applied for extracting the spectral features. The acquired spectral features are considered into dimension reduction phase using principle component analysis (PCA). Finally, the dimensionality reduced features are subjected into the optimized deep neural network (DNN), in which the parameters are tuned by Horse Herd Optimization (HHO). Experimental analysis reveals that the efficiency of proposed face recognition model is higher than the conventional models.