Emotion-Based Music Recommendation System- A Deep Learning Approach

People and musicians often form a close bond, some songs can transport listeners to different eras or locations, and listeners frequently turn to music for emotional support. By classifying music according to emotions, emotion-based recommendation algorithms and music suggestion assist users in elevating their present mood. This also saves time, as it suggests music based on the user's mood rather than forcing them to perform a time-consuming search. Aural feature categorization, wearable computers, and hand-performed music are just a few of the methods used these days. Most of these methods were developed using CNN and VGG-16 models to improve accuracy and performance, these models have produced highest accuracy rates of 68.28% and 54.27%, respectively. Thus far, we have suggested a method that leverages big data analytics and deep learning models to improve accuracy rate and performance. Through this paradigm, which employs streamlit to deliver an intuitive online interface, customers can access services. This paper compares assessment metrics of deep learning models trained and assessed on the FER2013 and CK+48 datasets on CNN, VGG with XGBoost, and VGG-16. For photos taken in static mode, the accuracy of the emotion extraction method ranges from 79% to 98%. On the CK+48 and Fer2013 datasets, VGG with XGBoost receives scores of 98% and 85%, respectively. Facial expressions are captured by an integrated webcam, and features are extracted to identify a range of emotions, including happy, disguist, angry, contempt, sadness, surprise, and fear. The YouTube page is redirected to offer a recommended music playlist. This work focuses on the reader's emotions while highlighting the importance of music. Here's a brief explanation of the idea.