Visual Emotion Recognition Through Affective Computing

Affective computing is currently one of the most active research domains which include speech descriptors, facial affect detection, multimodal system, and emotion classification. Affective computing is a multidisciplinary knowledge, such as psychology, cognition, and computer sciences. We present an overview of affective computing emotion theories, state of the art of key technologies, projects, research challenges, emotion feature extraction, algorithms, and applications. This paper analyzes the various emotions of humans such as happy, sad, anger, surprise, and neutral emotions based on the Warsaw set of emotional facial dataset. This research work comprises two phases. The first phase contains the preprocessing of given image data. The preprocessing consists of grayscale conversion, histogram equalization, face detection algorithm, and image cropping and resizing. The second stage is classification of emotions based on the 3D-CNN based SE-Net approach. Our proposed work attains the recognition accuracy of 95%. The performance measures of the research work have been evaluated. The performance metrics are precision, recall, and confusion matrix.