RhythmQuest: Unifying Indian Music Classification and Prediction with Hybrid Deep Learning Techniques
Priyanka Kaushik, Saurabh Pratap Singh Rathore, Kartik Chahal, Sparsh Saraf, Ganga Singh Chauhan, Pawan Kumar
Abstract
Indian music is a diverse and culturally rich art form that encompasses various genres, ragas, and styles. It accurately classifying and predicting Indian music compositions poses a significant challenge due to the complexity and nuances involved. Mostly recently, DL techniques have shown remarkable success in music classification tasks. This paper presents a study on Indian Music Classification and Prediction using a Hybrid DL Approach, which combines the strengths of DL models, such as RNN, LSTM, GRU, CNN, AlexNet, and ResNet, with traditional approaches. The proposed methodology involves collecting diverse Indian music datasets, preprocessing the data, creating annotations, and training hybrid models. Evaluation metrics and simulation results are used to assess the performance of the models, and comparative analysis is performed to demonstrate the superiority of the hybrid approach over traditional methods. The paper also discusses future directions, including the incorporation of contextual information, multimodal approaches, and improving interpretability and explain ability. Overcoming challenges such as data sparsity, domain adaptation, and leveraging expert annotations are also highlighted. This paper contributes to the field of Indian music classification and prediction, providing insights into the potential of hybrid DL approaches in unlocking the intricate patterns and structures of Indian music compositions.