An Analysis of Recommendation System Based on Multimodal Emotion Detection Using Facial Expression and Voice Modulation

The human subject's emotions and mood swings allow us to determine whether or not they are in good psychological health. The most glaring flaw in today's emotion recognition systems was their inability to interpret people's emotions through facial expressions, especially in cases of mental health issues. Depression is fast emerging as a leading health issue; a major hurdle in its treatment is lack of detection. In addition, many times persons who are in a particular mood are clueless about how to increase or decrease the emotions that go along with it. Individuals convey their sentiments and thoughts at the same time by using vocal and facial indicators. In this work, we examine emotion detection approaches, recommendations, research gaps, challenges, and future directions in accordance with the findings. Conventional techniques for emotion recognition emphasize uni-modality and mostly rely on manually created features. But for difficult emotion recognition tasks, these methods are unable to extract enough discriminative information. In order to address this problem, deep neural model-based methods and multimodal approaches have become more and more common in emotion recognition tasks. Subsequently, the recommendation system may suggest delightful activities that improve an individual's mental health. Recent years have seen a huge increase in interest in emotion recognition because of its important applications to human-computer interaction. The research community is very interested in FER and SER because of the recent advancements in deep learning techniques. Therefore, it is possible to employ AI or ML to identify a person's emotions and then provide an appropriate suggestion. Main goal is to assess people's emotions and recommend beneficial activities that will improve their mood and maintain good mental health.