Course Project for Music Information Retrieval at UPF Barcelona 2018-19.
Tabla bols were sampled from a dataset of Indian Tabla bols available at https://compmusic.upf.edu/tabla-solo-dataset
Experiments were performed with different machine learning algorithms to train a classifier model to recognize the tabla bol using a variety of audio features extracted using Essentia library.
Performance of around 72% classification accuracy was obtained using the HTK-MFCC features with an SVM classifier.
All the code is contained in the jupyter notebook MIR_Submission.ipynb
All packages used in the code can be installed using pip3 install <package-name>