Facial Emotion Recognition-Based Music System Using Convolutional Neural Networks

Abstract

Another type of conversational AI is the Facial Emotion Recognition (FER) that has recently been catapulte to mainstream use owing to its versatility since it is applicable in improving user experience of nearly all applications and services. In this respect, FER can help explain how to turn known people’s feelings into a higher level of appreciation to help improve user satisfaction. This work proposes an intricate architecture that is powered by CNN for facial emotions recognition, with the resultant generation of playlists with similar emotions. It is assumed that by analyzing real-time facial data received via video streams and carrying out further sophisticated emotion determination, the system tries to provide an auditory experience as close as possible to the user’s mood at the time of an event. This paper’s proposed system combines powerful data preprocessing, optimization of the CNN architectures, and a well-functioning emotion to music mapping system making it effective and easily scalable. Based on different datasets and realistic experiments, we have demonstrated the potential and efficiency of the presented framework. The system shows a good level of success in identifying affective states and suggests the pieces of music that are perceived by users as emotionally consonant, which confirms the possibility of using the method in practical tasks.