An Efficient Multimodal Emotion Identification Using FOX Optimized Double Deep Q-Learning

The physiological process of emotion is a comprehensive human state brought on by the conscious and unconscious observation of circumstances connected with various aspects like motivation, personality, and mood. In daily life, emotions are more important while making decisions, communicating, etc. The state-of-art techniques often classified different types of emotions such as sadness, anger, happiness, surprise, disgust, and hatred by analyzing the emotions present in the text. Very few works have focused on identifying human emotions by analyzing different physiological signals obtained using the Electrocardiogram (ECG), Electroencephalogram (EEG), and Galvanic Skin Response (GSR) modals. This paper mainly explores different human emotions using the physiological signals obtained from the ECG, EEG, and GSR modalities. In this paper, we propose a FOX-optimized Double Deep Q-learning (DDQ) model for identifying multimodal emotions using the signals such as GSR, ECG, and EEG. Here, the created Q value for multimodal emotion recognition is obtained using the Q-learning technique and used for training the parameters. The target Q-network is used in place of the Q-learning model to generate target Q values for training parameters. The DDQ hyperparameters and the local optima is balanced using the FOX algorithm. The performance of the proposed technique is evaluated in terms of F1 score and accuracy for predicting different sentiments through experiments conducted utilizing the ASCERTAIN database. According to the simulation results, the approach had a greater level of accuracy when identifying the five emotions of arousal, valence, engagement, liking, and familiarity. Also, the proposed approach has a lower standard deviation. Therefore, the proposed method is effective at identifying five human emotions.