Facial Emotion Based Content Recommendation

Abstract

Current state of emotion of a person is highly related to what entertainment content he/she want to listen or watch. An emotion-based content recommendation will help the user to not only to get content according to their current state of mind but also reduce the efforts of managing a playlist for music and help them reduce their stress level by recommending them appropriate content for stress relief. Emotion of a person can be determined using his/her facial expression. This facial expression can be detected using a machine learning model, we have developed a model using xception architecture. An application which will access the camera of the device and take image of the persons face, it will connect to the ML Kit stored on the cloud (Firebase) which will analyze the image and detect the mood of the user, from that mood it will connect to API of a music and movies application (E.g., Spotify, Netflix, Disney Hotstar, etc.), though which we will recommend the content. The application will also verify from the user for his/her taste of music and customize the recommendation accordingly. The user will be prompted for change in emotion after specific intervals if he/she likes to change the content.