Recommending Music tracks based on Listener’s Emotional State using various Architectures

Everyone in the modern era is influenced by stress. Many health issues are developing as a result of stress. Most people are spending more money and undergoing treatment to lessen their stress. One of the best strategies for lowering stress is to listen to music. Therefore, it is essential to create a recommender system that could create personalized music collection using machine learning (ML) and deep learning (DL) algorithms based on the user’s current mood captured through a web camera. In the present popular artificial intelligence (AI) field, recognizing an individual’s emotions based on their facial expression is very much essential. The idea behind this paper is to recognise music and help the user by recognizing their emotions based on their facial expressions as music and emotion are strongly correlated. Music recommender systems (MRS) act as decision support systems that lessen information overload by only obtaining the content that is thought to be useful to listeners based on their predicted moods. The objective of this study is to perform a comparative study between five deep network architectures. The highest accuracy of 89.16% is achieved by Mobilenet architecture while the lowest accuracy of 85.81% is achieved by VGG16 architectures. Further, a music playlist is generated according to the emotion of the user using real-time detection using the most effective architecture.