ARCHIVES

The interpretation of human faces is essential in human computer interface, affective computer interface and behavioral computing. The vast majority of conventional facial emotion recognition theories are based on classifying the expression into one dominant emotion which does not necessarily reflect the complexity of emotions in the real world. The nature of human feelings is ambivalent and is often manifested as a mixture of different emotional states together with different degrees of intensity. The drawback inspires the desire to have a more expressive and lifelike emotion recognition system.The current paper is a Convolutional Neural Network (CNN) based emotion distribution learner of facial images. It uses a CNN model to extract discriminative facial features by first doing face detection, alignment, normalization, and noise reduction. Rather than making a prediction that is a single emotion label, the model provides a probability distribution among more than two emotion categories, including emotional intensity and coexistence. It is suggested that the approach will enhance the realism and interpretability of emotion analysis and will effectively be used in mental health monitoring, social robotics, virtual assistants, and user experience analysis