Hybrid hunt-based deep convolutional neural network for emotion recognition using EEG signals

Emotion recognition from the electroencephalogram (EEG) signals is a recent trend as EEG generated directly from the human brain is considered an effective modality for recognizing emotions. Though there are many methods to address the challenge associated with the recognition, the research community still focuses on advanced methods, like deep learning and optimization, to acquire effective emotion recognition. Hence, this research focuses on developing a well-adapted emotion recognition model with the aid of an optimized deep convolutional neural network (Deep CNN). The significance of this research relies on the proposed hybrid hunt optimization, which engages in selecting the informative electrodes based on the neuronal activities and tuning the hyper-parameters of Deep CNN. Moreover, the frequency bands are analyzed, and frequency-based features are utilized for emotion recognition, which further boosts the recognition efficiency, increasing the significance of EEG as an accurate modality for recognizing emotions. The analysis is done using the DEAP and SEED-IV datasets based on performance parameters, such as accuracy, specificity and sensitivity, and the frequency bands. The accuracy of the proposed recognition model is 96.68% using the DEAP dataset concerning the training percentage and 95.89% using the SEED-IV dataset concerning the k-fold.