Multi-modal with Multiple Image Filters for Facial Emotion Recognition

Abstract

The need to understand people, especially their behaviors and feelings, is growing significantly in today’s quickly-moving world. Despite the remarkable progress of science and technology in general and artificial intelligence in particular, facial emotion recognition remains challenging. This paper proposes a unique method for enhancing the accuracy of emotion recognition models. Through image analysis, the hair area and other facial areas have similar pixels but different intensities. However, to recognize emotions on the face, people only need to focus on facial features. Therefore, areas with the same pixels are not very helpful in accurately recognizing emotions. To solve the above problem, we conducted to eliminate or blur pixels that are the same as on the facial image. In addition, we also demonstrate that using the multi-model approach can support the learning process by allowing the sub-models to collaborate and increase accuracy. The experiments showed that the proposed approach offers a valuable contribution to the field of facial emotion recognition and has a significant improvement compared to previous approaches.