Music Genre Classification with LSTM based on Time and Frequency Domain Features

Deep features generated from deep learning models contain more information for music classification than short-term features. This paper uses a long-short term memory (LSTM) model to generate deep features and achieve music genre classification. Firstly, two short-term features of Zero crossing rate (ZCR) and mel-frequency spectral coefficients (MFCC) are extracted from music in digital form, which is a time-domain feature and frequency-domain feature, respectively. Then these two features are fed to LSTM to generate deep features. Finally, we use support vector machine (SVM) and k-nearest neighbors (KNN) respectively to classify the music genre based on these deep features. Experimental results show that using LSTM can significantly increase the accuracy of music genre classification.

I. Introduction

Listening to music is a common recreational activity in people's daily life. With the development of communication and network storage technology, large amounts of music are accessible to the general public. To better cater to listeners' preference, a new trend of effectively and efficiently retrieving music according to personal music interests is generated. One solution is to classify and assign labels to music based on genres, mood, artist, etc. [1]. In the early stage, the work of music classification is done manually. However, manual classification can be wrong and cannot provide large scale recommendation [2]. Now music companies are trying to solve this problem using efficient machine learning models. On the one hand, it can reduce the cost of managing and maintaining digital music collections because thousands of music needs to be uploaded to music libraries every year. On the other hand, the classification model can help users find music they are interested in and provide effective recommendation service.