Study and analysis of feature based automatic music genre classification using Gaussian mixture model
Chandanpreet Kaur, Ravi Kumar · 2017 International Conference on Inventive Computing and Informatics (ICICI) · 2017
Music genre classification is a very popular tool used extensively in music industry all over the world. The information of rhythmic structure, instrumentation, and harmonic content about the music signal is visualized with the aid of genre. The task of genre classification was performed manually in early days. With the advancements in technology and signal processing algorithms, this task can now be done automatically. This paper reports the genre classification of audio signal by using multiple feature sets. Mel-Frequency Cesptral Coefficients (MFCCs), Spectral Rolloff, Time Domain Zero Crossings and Flux are the features used in this work. Gaussian mixture model (GMM) based classifier is employed for the final classification task. Simulation results are provided based on standard genre database using MATLAB. It is observed that the percentage classification improves when more number of features are used.