Unlock instant, AI-driven research and patent intelligence for your innovation.

Melody generation method based on generative adversarial network

A melody and network technology, applied in the field of automatic music melody creation, can solve the problems of inability to guarantee music quality requirements, lack of diversification, etc., and achieve the effect of shortening training time and increasing judgment

Active Publication Date: 2019-04-05
成都潜在人工智能科技有限公司
View PDF13 Cites 14 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The music generated by using this method of line drawing, after accumulating a certain amount, will produce music with a high degree of similarity in probability, without diversification
And it cannot guarantee the quality requirements of the music generated each time

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Melody generation method based on generative adversarial network
  • Melody generation method based on generative adversarial network
  • Melody generation method based on generative adversarial network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0064] In one embodiment of the present invention, the melody is generated by using deep convolution generation confrontation network (DCGAN) training, first selects a MIDI file of a style, and the style can be pop music, classical music, rock music, etc., and then take the note pitch as the vertical Coordinates, establish a coordinate system with time as the abscissa, and express one of the melody tracks of the MIDI file in the coordinate system; set the time step, assuming that the 64th note is a time step, then the time step of the 64th note is 1 , the time step of the 32nd note is 2, the time step of the 16th note is 4, and so on. Divide the melody track in the coordinate system into multiple time steps, and each time step is an event; set the event number according to the pitch of the note; count all the event numbers of the melody track, and get the event sequence .

[0065] Taking the event sequence as the real data r1 of DCGAN, the generated data x1 of the generator, f...

Embodiment 2

[0068] In the second embodiment of the present invention, adopt WassersteinGAN (WGAN) to train and generate melody, first select the MIDI file of a style, then take the note pitch as the ordinate, take the time as the abscissa to set up a coordinate system, the MIDI file One of the melody tracks is represented in the coordinate system; set the time step, assuming that the 64th note is a time step, the time step of the 64th note is 1, the time step of the 32nd note is 2, and the time of the 16th note The step is 4, and so on. Divide the melody track in the coordinate system into multiple time steps, and each time step is an event; set the event number according to the pitch of the note; count all the event numbers of the melody track, and get the event sequence .

[0069] Taking the event sequence as the real data r1 of WGAN, the generated data x1 of the generator, for the discriminator, its label determines that the output of r1 should be "1" normally, and the output of x1 sho...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a melody generation method based on a generative adversarial network. The melody generation method comprises the following steps: data pretreatment: extracting an event sequence of a melody track from an original MIDI file; generator training: training the event sequence by applying the generative adversarial network to obtain a GAN generator model; music generation: generating music by utilizing the GAN generator model. The melody generation method based on the generative adversarial network, disclosed by the invention, has the beneficial effects that a melody is generated by adopting the generative adversarial network; through the remarkable characteristics of a generative adversarial model, namely the characteristic of continuously performing game optimization between generation and discrimination, the high-quality melody is obtained to help a composer to generate the original melody and be conducive to artistic creation; an enhanced discriminator is additionally arranged at the front part of a discriminator to increase the judgment of domain knowledge, therefore, the rapid convergence iteration of the discriminator in the training process is more facilitated, and the training time is shortened.

Description

technical field [0001] The invention belongs to the technical field of automatically creating music melodies, and in particular relates to a method for generating melodies based on generative confrontation networks. Background technique [0002] Melody is the foundation of music. Whether it is the "Book of Songs", the beginning of ancient Chinese poetry, or the current popular music, it is inseparable from the melody when it is performed. convey emotion. Melody is an important part of music, and the creation of melody is also related to the quality of music. Traditional composition requires the composer to have a certain knowledge of music theory, combined with inspiration and creative experience, in order to create a complete musical melody. [0003] With the development of computer technology, there are more and more computer-based assistant creation tools. The invention patent with the publication number CN104485101B discloses a method for automatically generating musi...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G10H1/00
CPCG10H1/0033G10H1/0091G10H2210/101G10H2210/005
Inventor 尹学渊陈洪宇陈超
Owner 成都潜在人工智能科技有限公司