Publication

Abstract : Present-day listeners can find it challenging to invest the time and effort necessary to learn the fundamentals of a particular genre of music. It would be easier to study Carnatic music in a shorter length of time if a paradigm that classifies the ragas existed. This would appeal to a bigger number of people since it would take less time to learn Carnatic music. In this sort of exploratory study, audio recordings are utilized as input to construct a model that can categorize classical music into a range of ragas. The model can also classify different types of classical music. For this, the features are retrieved by employing signal processing techniques to extract MFCCs, Spectral Rolloff, Zero Crossing rate, Spectrograms, and Spectral Centroids. After that, RNN Deep Learning models viz. LSTM and Bidirectional LSTM were trained and assessed to evaluate the model performance. The evaluation accuracy of these models came out to be 92.13% and 97.63% respectively.