Abstract
AbstractMusical genres are used as a way to categorize and describe different types of music. The common qualities of musical genre members are connected with the song's melody, rhythmic structure, and harmonic composition. The enormous amounts of music available on the Internet are structured using genre hierarchies. Currently, the manual annotation of musical genres is ongoing. Development of a framework is essential for automatic classification and analysis of music genre which can replace human in the process making it more valuable for musical retrieval systems. In this paper, a model is developed which categorizes audio data in to musical genre automatically. Three models such as neural network, CNN, and RNN-LSTM model were implemented and CNN model outperformed others, with training data accuracy of 72.6% and test data accuracy of 66.7%. To evaluate the performance and relative importance of the proposed features and statistical pattern recognition, classifiers are trained considering features such as timbral texture, rhythmic content, pitch content, and real-world collections. The techniques for classification are provided for both whole files and real-time frames. For ten musical genres, the recommended feature sets result in a classification rate with an accuracy of 61%.KeywordsMusic genre classificationMachine learningDeep learningMel frequency coefficient of cepstrum (MFCC)Convolutional neural network (CNN)Recurrent neural network-long short-term memory (RNN-LSTM) network
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call