Rāga Recognition in Indian Carnatic Music Using Transfer Learning
S. Sreejith, Rajeev Rajan
Abstract
An important aspect of Indian Classical music (ICM) is rāga, which serves as a melodic framework for both traditions of classical music. In this work, we propose a CNN-based sliding window analysis on Mel-spectrogram for rāga recognition in Carnatic music. The important contribution of the work is that the proposed method neither requires pitch ex-traction nor metadata for the estimation of rāga. CNN learns the representation of rāga from the patterns in the Mel-spectrogram during training through a sliding-window analysis. We train and test the network on sliced-Mel-spectrogram of the original audio while the final decision is made on the audio as a whole. The performance is evaluated on 10 rāga s from the CompMusic dataset and the Ramanarunachalam carnatic youtube collection. Out of the two proposed schemes, aggregation-based VGG16 model reports a macro F1 measure of 0.61, which is comparable to the result obtained for the base-line sequence classification model.