Modification of codewords in dictionary used for efficient coding of digital media spectral data

• US 20070016414A1
• Filed: 07/15/2005
• Published: 01/18/2007
• Est. Priority Date: 07/15/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Chen, Wei-Ge, Koishida, Kazuhito, Mehrotra, Sanjeev
• Granted Patent: US 7,562,021 B2
• Time in Patent Office: Days
• US Class Current: 704/230
• CPC Class Codes: G10L 19/038 Vector quantisation, e.g. TwinVQ audio
• CPC Class Codes: G10L 19/24 Variable rate codecs, e.g. for generating different qualities using a scalable representation such as hierarchical encoding or layered encoding

Abstract

Coding of spectral data by representing certain portions of the spectral data as a scaled version of a code-vector, where the code-vector is chosen from either a fixed predetermined codebook or a codebook taken from a baseband. Various optional features are described for modifying the code-vectors in the codebook according to some rules which allow the code-vector to better represent the data they are modeling. The code-vector modification comprises a linear or non-linear transform of one or more code-vectors, such as, by exponentiation, negation, reversing, or combining elements from plural code-vectors.

Claims

1. An audio encoding method, comprising:

providing codewords comprising a library of codewords;

transforming at least one codeword from the library;

comparing a sub-band to at least one transformed codewords from the library;

coding the sub-band in an output bitstream comprising coding an identifier of one or more codewords from the library and a transform identifier.

2. The encoder of claim 1 further comprising:

transforming an input audio signal into a set of spectral coefficients;

coding a baseband portion of the set of spectral coefficients in the output bitstream;

dividing an extended band of the spectral coefficients into plural sub-bands;

scaling the plural sub-bands in the extended band; and

comparing the sub-band to at least one codeword from the library that has not been transformed, wherein the library comprises plural codewords from the baseband portion.

3. The audio encoding method of claim 1 wherein available transforms for transforming at least one codeword from the library comprise one or more of the following transforms:

applying an exponent to each coefficient of a codeword;

negating each coefficient of a codeword;

or

reversing the order of coefficients in a codeword.

4. The audio encoding method of claim 1 wherein transforming at least one codeword from the library comprises creating a codeword with coefficients from two or more codewords comprising:

from all but the final codeword, selecting coefficients that satisfy a rule;

from a final codeword, providing the other coefficients.

6. The audio encoding method of claim 1 wherein the library further comprises codewords from a noise codebook or a codeword populated using a determinatively seeded random number generator.

7. The audio encoding method of claim 1 wherein coding the sub-band includes providing an identifier of two or more codewords and the transform identifier comprises at least one of an exponent indication, a sign indication, a direction indication, or an ordering of codeword identifiers in the output bitstream, the ordering indicating an implicit selection of coefficients.

8. The audio encoding method of claim 1 wherein coding the sub-band in the output bitstream includes an identifier of two or more codewords and the transform identifier is an identifier of an explicit rule for selection of coefficients from the two or more codewords.

9. The audio encoding method of claim 1 wherein the compared at least one transformed codeword from the library is two or more codewords created using an exponential transformation of a closest matching codeword from the library.

10. The audio encoding method of claim 9 wherein the closest matching codeword from the library is identified using a least-mean square comparison and the two or more codewords created from the exponential transformation are compared using a probability mass function.

11. The audio encoding method of claim 1 wherein the compared codewords comprise plural codewords from the library and comparing the sub-band to the at least one transformed codeword from the library comprises an exhaustive search on the codewords of the library and transformations thereof comprising negation, reverse direction, and exponential transformations using two or more exponents.

12. The audio encoding method of claim 2, further comprising:

determining that a part of the baseband portion poorly represents the input audio signal;

enhancing the part of the baseband portion;

the enhancement comprising, from the poorly represented part of the baseband portion, selecting coefficients that represent the input audio signal well, and from a second codeword, selecting all other coefficients; and

coding the enhancement comprising an identifier of the second codeword, an identifier of the poorly represented part, and a rule for selecting coefficients.

13. The audio encoding method of claim 12 wherein the second codeword is obtained from a noise codebook or random number generator.

14. The audio encoding method of claim 1 wherein transforming at least one codeword from the library comprises creating a codeword with coefficients from two or more codewords comprising:

from a first codeword, selecting coefficients that satisfy a rule; and

for coefficients in the first codeword that do not satisfy the rule, performing a mathematical operation to create other coefficients, the mathematical operation comprising an operator and plural operands,

a first operand being a coefficient from the first codeword that does not satisfy the rule, and

a second operand being a coefficient obtained from a second codeword.

15. The audio encoding method of claim 1, further comprising pre-selecting codewords before comparing the sub-band to codewords, the pre-selection comprising:

creating an envelope comprising running a weighted average function on an audio signal; and

determining the pre-selected codewords by comparing the envelope to the sub-band.

16. The audio encoding method of claim 15 wherein comparing the envelope to the sub-band further comprises:

transforming the envelope using one or more transforms comprising a negation transform, a reverse transform, or an exponential transform; and

wherein comparing the envelope to the sub-band comprises determining a Euclidean distance.

5. (canceled)

17. An audio decoding method comprising:

decoding encoded spectral coefficients in a bitstream; and

decoding one or more encoded sub-bands in the bitstream comprising,

determining one or more codeword identifiers for each sub-band,

obtaining the one or more determined codewords for each sub-band, and

for at least one sub-band, determining a transformation rule,

for the at least one sub-band, transforming a codeword obtained for the sub-band using the transformation rule.

18. The audio decoding method of claim 17 wherein the determined transformation rule comprises one or more of the following transforms:

applying an exponent to each coefficient of a codeword;

negating each coefficient of a codeword;

or

reversing the order of coefficients in a codeword.

19. The audio decoding method of claim 17 wherein the determined transformation rule creates a codeword from two or more codewords comprising:

from all but the final codeword, selecting coefficients that satisfy a rule; and

from a final codeword, providing the other coefficients.

20. An audio encoder comprising:

a transform for transforming an input audio signal block into spectral coefficients;

a base coder for coding values of a baseband portion of spectral coefficients into a bitstream;

a divider for dividing a portion of spectral coefficients into sub-bands;

a scaler for scaling sub-bands;

a comparer for comparing sub-bands to codewords from a library of codewords;

an extended band coder for coding sub-bands into the bitstream, wherein a

coded sub-band comprises an identifier of a codeword and a exponent for transforming the identified codeword.