Abstract:

Voice analysis is a type of analysis that can be applied in many areas such as disease detection, emotion analysis, and classification of music genres. Classification can be made by applying image or data analysis according to the problem with voice characteristics. In this study, the classification of music genres was performed by extracting the voice characteristics. GTZAN, which consists of 10 music genre tags, was used as the dataset. In the analysis phase, the effects of audio segmentation and feature selection methods on the classification of music genres were investigated. Machine learning methods and deep neural networks were used for classification. In the first stage of the analysis, the accuracy value increased by 10.58% by applying only the audio segmentation. As a result of this study, after the application of audio segmentation and feature selection methods, Deep Neural Network provided an increase of 14.19% with an accuracy of 95.19%.

Keywords: