Creative GAN generating music deviating from style norms

Patent metadata

Field	Value
Publication number	US-11250608-B2
Application number	US-202017108432-A
Country	US
Kind code	B2
Filing date	Dec 1, 2020
Priority date	Jun 20, 2018
Publication date	Feb 15, 2022
Grant date	Feb 15, 2022

Abstract

A method and system for generating music uses artificial intelligence to analyze existing musical compositions and then creates a musical composition that deviates from the learned styles. Known musical compositions created by humans are presented in digitized form along with a style designator to a computer for analysis, including recognition of musical elements and association of particular styles. A music generator generates a draft musical composition for similar analysis by the computer. The computer ranks such draft musical composition for correlation with known musical elements and known styles. The music generator modifies the draft musical composition using an iterative process until the resulting musical composition is recognizable as music but is distinctive in style.

First claim

I claim:

1. A method for generating music comprising the steps of:
a) providing to a computer a set of digitized data characterizing a plurality of existing musical compositions created by human beings, the digitized data including the designation of a style associated with each such existing musical composition, the computer including a memory;
b) using the computer to analyze the digitized data and corresponding style designations, including recognition of musical elements in the existing musical compositions, and creating music recognition data associating particular existing musical compositions with particular style designations;
c) storing the music recognition data in the memory of the computer;
d) providing a music generator to generate a musical composition;
e) providing the generated musical composition generated by the music generator to the computer;
f) using the computer to compare the generated musical composition to the stored music recognition data;
g) based upon such comparison, providing a music index which ranks the generated musical composition along a scale that ranges from including existing musical elements found in existing musical compositions and not including existing musical elements found in existing musical compositions;
h) based upon such comparison, providing a novelty index which ranks the generated musical composition along a scale that ranges from being within a style associated with existing musical compositions and not being within a style associated with existing musical compositions;
i) modifying the generated musical composition based upon the music index and the novelty index;
j) repeating steps e) through i) until the generated musical composition produces a music index indicating that the generated musical composition includes existing musical elements found in existing musical compositions, while also producing a novelty index indicating the generated musical composition is not within a style associated with existing musical compositions.

2. The method recited by claim 1 wherein such method is an iterative enhancement process.

3. The method recited by claim 1 wherein the step of using the computer to analyze the digitized data includes the use of a neural network.

4. The method recited by claim 1 wherein the step of modifying the generated musical composition based upon the music index and the novelty index includes the step of providing the music index and the novelty index to the music generator.

5. The method recited by claim 1 including the additional step of providing a random input vector to the music generator to generate a series of random musical elements.

6. The method recited by claim 1 wherein a musical composition is digitized and encoded using frequency analysis.

7. The method recited by claim 1 wherein a musical composition is digitized and encoded using music transcription methods.

8. The method recited by claim 1 wherein the step of using the computer to compare the generated musical composition to the stored music recognition data includes configuring the computer to provide a discriminator.

9. The method recited by claim 8 wherein the discriminator receives digitized data representing known musical compositions for training purposes, and the discriminator receives digitized data of a generated musical composition.

10. A computerized system for generating music comprising in combination:
a) a computer for receiving a set of digitized data characterizing a plurality of existing musical compositions created by human beings, the digitized data including the designation of a style associated with each such existing musical composition;
b) an electronic memory coupled to the computer;
c) the computer analyzing the digitized data and corresponding style designations, including recognition of musical elements in the existing musical composition, and creating music recognition data associating particular artistic elements of existing musical compositions with particular style designations for storage in the electronic memory;
d) a music generator coupled to the computer for generating a musical composition, and presenting such generated musical composition to the computer;
e) the computer comparing the generated musical composition to the stored music recognition data, and the computer providing a music index which ranks the generated musical composition along a scale that ranges from including existing musical elements found in existing musical compositions and not including existing musical elements found in existing musical compositions, the computer also providing a novelty index which ranks the generated musical composition along a scale that ranges from being within a style associated with existing musical compositions and not being within a style associated with existing musical compositions;
f) the music generator receiving information associated with the music index and the novelty index, and being responsive thereto to generate a modified musical composition for presentation to the computer;
whereby the music generator repeatedly presents modified musical compositions to the computer until a final modified musical composition presented thereby to the computer produces a music index indicating that the generated musical composition includes existing musical elements found in existing musical compositions, while also producing a novelty index indicating the generated musical composition is not within a style associated with existing musical compositions.

11. The computerized system recited by claim 10 wherein the computer is configured to form a neural network to analyze the digitized data.

12. The computerized system recited by claim 10 further including a random input vector source coupled to the music generator to generate a series of random musical elements.

13. The computerized system recited by claim 10 wherein a musical composition is digitized and encoded by the computer using frequency analysis.

14. The computerized system recited by claim 10 wherein a musical composition is digitized and encoded by the computer using music transcription methods.

15. The computerized system recited by claim 10 wherein the computer is configured to form a discriminator to compare the generated musical composition to the stored music recognition data.

16. The computerized system recited by claim 15 wherein the discriminator receives digitized data representing known musical compositions for training purposes, and the discriminator receives digitized data of a generated musical composition.

17. A method for generating music comprising the steps of:
a) providing to a computer a set of digitized data characterizing a plurality of elements of existing music, the digitized data including the designation of a style associated with each such element of existing music, the computer including a memory;
b) using the computer to analyze the digitized data and corresponding style designations, including recognition of elements of existing music, and creating music recognition data associating particular existing music elements with particular style designations;
c) storing the music recognition data in the memory of the computer
d) providing a music synthesizer to generate music elements;
e) providing the generated music elements generated by the music synthesizer to the computer;
f) using the computer to compare the generated music elements to the stored music recognition data;
g) based upon such comparison, providing a music index which ranks the generated music elements along a scale that ranges from in

Assignees

• Elgammal Ahmed
• Univ Rutgers

Inventors

• Elgammal Ahmed

Classifications

• G10H1/0025 Primary

  Automatic or semi-automatic music composition, e.g. producing random music, applying rules from music theory or modifying a musical piece (automatically producing a series of tones G10H1/26) · CPC title

• G06V10/7753

  Incorporation of unlabelled data, e.g. multiple instance learning [MIL] · CPC title

• G06V20/00

  Scenes; Scene-specific elements (control of digital cameras H04N23/60) · CPC title

• G06V10/82

  using neural networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title