IRJET- Developing a Framework for Music Genre Classification using Machine Learning Algorithms (original) (raw)
Due to the rapid variants in music tracks, music genre classification plays an intriguing part in today's globe. We'd want to index them in order to have better access to them. Music genres enable access to a huge range of music. To simplify its architecture and index the same with the preference of the user, Machine Learning (ML) is used in the majority of contemporary music genre categorization systems. Hence, in this paper, we present a music dataset which includes ten different genres consisting of 100 samples of each genre. Different Machine Learning approaches are used to train and classify the system such as Decision Tree Classifier (DTC), Random Forest (RF), K-Nearest Neighbor (KNN) and Support Vector Machine (SVM). The performance of the classifiers is compared and evaluated based on certain evaluation parameters. Feature Extraction is the most crucial process for audio analysis. After research we choose Mel Frequency Cepstral Coefficient (MFCC), Pitch and Tempo as the required feature vector for the audio samples. Each algorithm is tested with a dataset of 1000 & 2000 music samples respectively. The suggested method categorizes music into several genres based on the feature vector characterization. The four algorithms used are based on supervised learning, where the output is predicted based on the training of the developed classifier. From the results, it is observed that the accuracy of genre prediction using RF is 97% & outperformed the rest when compared with state of the art for 2000 music samples.