Improved Music Genre Classification and Visualization using Conventional Neural Network

A genre in music is a style or kind of music. To make it easier to recognize favorite songs, music genres group related tracks together. Dissimilar machine learning approaches, for example Decision Tree, K-means Clustering, Support Vector etc., can be used to categorize diverse genres of music. This research proposes a genre classification system based on Conventional Neural Networks (CNNs). The acoustic signal is reformed into a spectrogram in the CNN model. Next, every audio file in a dataset has its Mel-Frequency Cepstral Coefficient (MFCC) retrieved, which is utilized to train the model. Any audio will be classified and its genre will be predicted by the developed model. Additionally, the suggested system is utilized to show the genres using various visualization techniques, including as stacked-bar charts, tables, pie charts, and stacked area charts. The audio is categorized using this system as Pop, Rock, Reggae, Hip-hop, Jazz, Blues, Country, Classical, and Disco. Using visualization techniques, the probability of each genre in an audio file are also displayed. Python is used to implement the suggested approach, and the model is educated by means of the GTZAN dataset. Pandas and matplotlib are used in the implementation of the visuals. On the GTZAN dataset, the precision of the suggested model is 0.875.