As the quantity of music being released on internet platforms is on a constant rise, genre classification becomes an important task, and has a wide variety of applications. Based on the current statistics, almost 60,000 songs are uploaded to Spotify every day and hence the need for reliable data required for search and storage purposes climbs in proportion. Several Machine Learning and Deep learning models are developed to solve this problem. Some are listed below:
Aim of this project is to compare above four most used models and find out which model gives us the accurate results.
Haggblade, Michael et al. “Music Genre Classification.” (2011).[1] wrote a paper where they investigate various MAchine learning algorithms and a neural network to classify the genre. There study was relied on single chracteristic which is Mel Frequency Cepstral Coefficient(MFCC). Later, Hoang, Lam. "Literature Review about Music Genre Classification." (2018). [2] discussed how we can classify music using two major Neural networks CNN and RNN on same dataset but with many more music features. These two papers inspire us to compare all these models reliaying on atleat 4 music features.
[1] Haggblade, M., Hong, Y., & Kao, K. (2011). Music Genre Classification.
[2] Hoang, L. (2018). Literature Review about Music Genre Classification.
[3] Agrawal, M., & Nandy, A. (2020). A Novel Multimodal Music Genre Classifier using Hierarchical Attention and Convolutional Neural Network.
[4] Mandel, M., & Ellis, D. (2006). Song-level features and svms for music classification. In In Proceedings of the 6th International Conference on Music Information Retrieval, ISMIR (Vol. 5).