Emotion Based Carnatic Music Recommendation System

The project aims to develop an AI-based music recommendation system focused on Melakarta ragas and emotional themes. In the field, previous work has focused on music recommendation systems based on various techniques, but limited research has been conducted specifically on Melakarta ragas and their emotional impact. Therefore, there is a gap to propose a new idea that leverages Melakarta ragas to provide personalized music recommendations aligned with users' emotional states. The objectives include preprocessing the dataset, classifying Melakarta ragas, mapping emotions to ragas, and implementing the recommendation system. The methodology involves identification of Navarasa (nine emotions) from camera input. Computer vision techniques are employed to analyze facial expressions and detect emotions such as Hasya, Karuna, Shringara, etc. These identified Navarasas are then mapped to corresponding emotional tags associated with Melakarta ragas. The recommendation system is developed to map the detected Navarasa from camera input to emotional tags and provide personalized music recommendations based on the Melakarta ragas associated with those emotions. The simulation tools used include programming languages like Python, machine learning libraries like scikit-learn and computer vision frameworks like OpenCV. The paper concentrates on majorly 6 rasas that maps with all the 72 melakarta ragas.