Parametric Analysis on Enhanced Facial Emotion Recognition with Weight Optimized Neural Network

This paper intends to design a novel FER model that includes four phases such as (i) Face Detection, (ii) Feature extraction, (iii) Dimension reduction, and (iv) Classification. Here, Viola Jones (VJ) method is deployed for face detection, which is the initial technique to offer better object detection at real-time. Subsequently, feature extraction is carried out by means of Local Binary Pattern (LBP), and Discrete Wavelet Transform (DWT). As the “curse of dimensionality” seems to be a major fact, the dimension reduction of features is done using Principal Component Analysis (PCA). Finally, the classification is carried out by means of Neural Network (NN) with the new training algorithm called Probability based-Bird Swarm Algorithm (P-BSA), by which the weights are optimized. The performance of the proposed algorithm is done by making the algorithmic analysis. More importantly, the positive integer ($U$) of the proposed algorithm is varied to certain values: 0.5, 1, 1.3, 1.5 and 1.8, respectively.