Music classification method based on deep learning

Abstract

The invention discloses a music classification method based on deep learning, and the method comprises the steps: converting an audio file of music into a Mel spectrogram, and generating a one-hot vector of a music label; respectively transmitting the Mel spectrogram of the music and the one-hot vector of the music label into a convolutional neural network and a recurrent neural network; obtainingfrequency domain and time domain feature vectors of music through a convolutional neural network, obtaining a music-label relationship vector through a recurrent neural network, mapping respective output vectors of the convolutional neural network and the recurrent neural network to the same dimension, and performing connecting and combining to form a music-label embedding vector; and transmitting the music-label embedding vector to a label prediction layer, wherein the output of the music-label embedding vector is the probability value of the music corresponding to each category label, and finally, selecting a plurality of category labels from all the category labels as the classification of the music according to the probability values. The invention belongs to the technical field of information, and can realize accurate prediction of music category labels based on numerous and staggered relationships of music categories.

Description

technical field [0001] The invention relates to a music classification method based on deep learning, which belongs to the field of information technology. Background technique [0002] Music is the most popular art form performed and listened to by billions of people every day. There are many genres of music, such as pop, classical, jazz, folk, etc. Each genre has different instruments, timbres, rhythms, beats, flows, and more. Music genre classification is one of the many branches of Music Information Retrieval (MIR), which can be used to perform other tasks on music data, such as beat tracking, music generation, recommendation systems, track separation and instrument recognition, etc. [0003] In the traditional music classification method, a song often only belongs to a certain category, and the category is single and the number is small. However, considering the variety and interlacing of music categories today, the same song may belong to multiple categories. Theref...

AI Technical Summary

Problems solved by technology

[0003] In the traditional music classification method, a song often only belongs to a certain category, and the category is single and the number is small

However, considering the variety and interlacing of music categories today, the same song may belong to multiple categories. Therefore, how to realize the accurate prediction of music category labels, so as to automatically identify multiple categories of music, has become a technology that technicians generally pay attention to. question

Images