Validating the Genre using Different Features and Deep Neural Network for Marathi Songs

Abstract

Here we make known the Deep Neural Network for Genre classification of music category or type classification. Deep Neural networks are widely used nowadays for very large applications. DNN is a neural network with many hidden layers as per the user's requirement. DNN gives better results as compared to other classification methods and neural networks due to its iterative nature. The features which are extracted from the music are Mel Frequency Cepstrum Coefficient (MFCC), Zero Crossing Rate (ZCR), STFT, spectral bandwidth, Chroma. With each feature, features are extracted and DNN model is validated against each feature. It is observed that MFCC gave better results for genre classification. Here we used the Marathi song dataset and GTZAN dataset for classification and are validated.