Emotion Recognition Via Deep Learning Based Fuzzy CapsuleNet

Emotions are composed of conscious logical responses to various situations. Physiological, cognitive, and behavioural changes produce these mental responses. In order to understand a person's emotional state, one crucial piece of visual information called facial expression recognition (FER) can be used. In recent days, deep learning has achieved significant success in a number of fields, including signal processing and emotion recognition. In this work, the Extended Cohn-Kanade dataset is used as an input. The Gaussian denoising method is used for pre-processing the image for eliminating the noise. The LDA is used for feature extraction for extracting the most relevant features from the pre-processed images. Finally, the proposed FCAP model processed the extracted features of the images and classified them into different emotions. The efficiency of this model was accomplished by accuracy, specificity and sensitivity. According to the experimental result, The proposed FCAP method achieves the overall accuracy of 98.48%. The proposed FCAP model improves overall accuracy by 0.98%, 1.90%, and 0.28% better than XeceptionNet, Edge computing-CNN and CNN respectively.