A deep learning approach for emotion classification through facial expressions from images based on optimal quantum convolutional neural network

Abstract

Communication plays a prominent role in every person newline s daily life. newlineNon verbal communication relies heavily on facial expressions to transmit newlinefeelings, behaviors, and motivations. The process of classifying the newlineexpressions on face images into categories such as fear, anger, surprise, newlinesadness, happiness, disgust, and so forth is known as Facial Expression newlineRecognition, or FER. FER is an essential component of Human Computer newlineInteraction (HCI) that makes it possible to interpret facial expressions based newlineon human cognition. It has several uses in the fields of robotics, marketing, e newlinelearning, security, mental health treatment, and surveillance among others. The newlinelow prediction accuracy of the current Face Expression Recognition (FER) newlinealgorithms is the biggest issue. Despite experiencing rapid growth, facial newlineexpression recognition (FER) remains challenging due to factors such as head newlinedeflection, partial occlusions in the facial region, and variations in newlineillumination. These challenges can impact face detection performance and newlineresult in decreased accuracy in FER. newlineTo address these issues, a methodology is proposed called Ensemble newlineConvolutional Recurrent Neural Network (CRNN) to improve the accuracy of newlinefacial expression prediction. The model integrates methods such as Gated newlineRecurrent Units (GRU), Long Short Term Memory (LSTM), and newlineConvolutional Neural Network (CNN) to enhance prediction outcomes, and newlineemploys an Adaptive Neuro-Fuzzy Inference System (ANFIS) for newlinecomprehensive analysis. The evaluation using Facial Recognition Dataset and newlineEMOTIC dataset demonstrates the model newline s effectiveness, achieving high newlineaccuracy, precision, F1-score, False Positive Rate, and True Positive Rate. newline