Method and apparatus for classifying mood of music at high speed

• US 20070107584A1
• Filed: 09/12/2006
• Published: 05/17/2007
• Est. Priority Date: 11/11/2005
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Zhu, Xuan, Kim, Ji, Shi, Yuan, Kim, Hyoung, Eom, Ki
• Granted Patent: US 7,582,823 B2
• Time in Patent Office: Days
• US Class Current: 84/612
• CPC Class Codes: G10H 1/0008 Associated control or indicating means
• CPC Class Codes: G10H 2210/076 for extraction of timing, tempo; Beat detection tempo display G10H2220/081; tempo control G10H2210/375
• CPC Class Codes: G10H 2240/085 Mood, i.e. generation, detection or selection of a particular emotional content or atmosphere in a musical piece
• CPC Class Codes: G10H 2250/225 MDCT [Modified discrete cosine transform], i.e. based on a DCT of overlapping data adaptive MDCT compression, e.g. ATRAC [adaptive transform acoustic coding] G10H2250/575
• CPC Class Codes: G11B 27/105 of operating discs
• CPC Class Codes: G11B 27/28 by using information signals recorded by the same method as the main recording G11B27/22 takes precedence

Abstract

A method and apparatus for classifying mood of music at high speed. The method includes: extracting a Modified Discrete Cosine Transformation-based timbre feature from a compressed domain of a music file; extracting a Modified Discrete Cosine Transformation-based tempo feature from the compressed domain of the music file; and classifying the mood of the music file based on the extracted timbre feature and the extracted tempo feature.

Claims

1. A method of classifying a mood of a music file, comprising:

extracting a Modified Discrete Cosine Transformation-based timbre feature from a compressed domain of a music file;

extracting a Modified Discrete Cosine Transformation-based tempo feature from the compressed domain of the music file; and

classifying a mood the of the music file based on the extracted timbre feature and the extracted tempo feature.

2. The method of claim 1, wherein the classifying the mood of the music file comprises:

classifying a genre of the music file based on the extracted timbre feature; and

reclassifying a category of the music file of the genre when uncertainty of a genre classification result is greater than a predetermined value.

3. The method of claim 2, wherein, in the reclassifying a category of the music file of the genre, the category of the music file of the genre is reclassified based on the extracted tempo feature.

4. The method of claim 3, further comprising merging reclassified categories of the music file into K music moods.

5. The method of claim 2, wherein, in the classifying a genre of the music file, the genre of the music file is classified into one of a sad, an exciting, a calm in classic, a calm in pop, a pleasant in pop, a pleasant in classic, and a pleasant in jazz genre.

6. The method of claim 5, wherein, in the reclassifying a category of the music file of the genre, the category of the music file classified into the pleasant in classic genre is reclassified into the calm and the pleasant in classic genres according to the extracted tempo feature, and the category of the music file classified into the pleasant in jazz genre is reclassified into the sad and the pleasant in jazz genres according to the extracted tempo feature.

7. The method of claim 1, further comprising displaying a result of the classifying according to mood classification reliability.

8. The method of claim 7, wherein the mood classification reliability is based on a maximum likelihood value of results of a Gaussian Mixture Model, and the displaying comprises:

arranging the classified music file in an order descending from high mood classification reliability of the music file; and

displaying the classified music file in a playlist according to the order.

9. The method of claim 1, wherein the extracting a Modified Discrete Cosine Transformation-based timbre feature from the compressed domain of the music file comprises:

extracting Modified Discrete Cosine Transformation coefficients by decoding a part of the music file;

selecting the Modified Discrete Cosine Transformation coefficients of a predetermined number of sub-bands from the extracted Modified Discrete Cosine Transformation coefficients; and

extracting a spectral centroid, a bandwidth, a rolloff, and a flux from the selected Modified Discrete Cosine Transformation coefficients.

10. The method of claim 1, wherein the extracting the Modified Discrete Cosine Transformation-based tempo feature from the compressed domain of the music file comprises:

extracting the Modified Discrete Cosine Transformation coefficients by decoding a part of the music file;

selecting the Modified Discrete Cosine Transformation coefficients of a predetermined number of sub-bands from the extracted Modified Discrete Cosine Transformation coefficients;

extracting a Modified Discrete Cosine Transformation modulation spectrum from the selected Modified Discrete Cosine Transformation coefficients by performing Discrete Fourier Transformation;

dividing the extracted Modified Discrete Cosine Transformation modulation spectrum into N sub-bands; and

extracting a sub-band peak, a sub-band valley, and a sub-band average from the divided sub-band.

11. The method of claim 1, wherein, in the classifying the mood of the music file, the music file is classified into a subdivided category by a Gaussian Mixture Model.

12. The method of claim 1, wherein, in the classifying the mood of the music file, the mood of the music file is classified into one of sad, calm, exciting, and pleasant.

13. A computer readable recording medium in which a program for executing the method of claim 1 is recorded.

14. An apparatus for classifying a mood of a music file, comprising:

a timbre extraction unit extracting a Modified Discrete Cosine Transformation-based timbre feature from a compressed domain of a music file;

a tempo extraction unit extracting a Modified Discrete Cosine Transformation-based tempo feature from the compressed domain of the music file; and

a mood classification unit classifying the mood of the music file based on the extracted timbre feature and the extracted tempo feature.

15. The apparatus of claim 14, wherein the mood classification unit comprises:

a first classification unit classifying a genre of the music file based on the extracted timbre feature; and

a second classification unit reclassifying a category of the music file based on the extracted tempo feature when uncertainty of a genre classification result is greater than a predetermined value.

16. The apparatus of claim 14, further comprising a display unit displaying a result of the classifying of the mood of the music file in an order according to mood classification reliability.

17. A method of improving a reliability of music mood classification, comprising:

extracting Modified Discrete Cosine Transformation-based timbre and tempo features from a compressed portion of a music file; and

classifying a mood of the music file by classifying a genre of the music file based on the extracted timbre feature, ascertaining whether the genre resulting from the genre classification has an uncertainty in excess of a threshold amount, and reclassifying a category of the music file of the genre when the uncertainty of the genre classification exceeds the threshold amount.

18. A computer readable recording medium in which a program for executing the method of claim 17 is recorded.