Flexible and scalable combined innovation codebook for use in CELP coder and decoder

• US 9,053,705 B2
• Filed: 04/11/2011
• Issued: 06/09/2015
• Est. Priority Date: 04/14/2010
• Status: Active Grant

• Current Assignee: VoiceAge EVS LLC (FIG LLC (d/b/a Fortress Investment Group LLC))
• Original Assignee: VoiceAge Corporation
• Inventors: Bessette, Bruno
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Le, Thuykhanh
• Application Number:
  US13/083,900
• Publication Number:
  US 20120089389A1
• Time in Patent Office: 1,520 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G10L 19/0212 using orthogonal transformation
• CPC Class Codes: G10L 19/09 Long term prediction, i.e. removing periodical redundancies, e.g. by using adaptive codebook or pitch predictor
• CPC Class Codes: G10L 19/12 the excitation function being a code excitation, e.g. in code excited linear prediction [CELP] vocoders
• 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

In a CELP coder, a combined innovation codebook coding device comprises a pre-quantizer of a first, adaptive-codebook excitation residual, and a CELP innovation-codebook search module responsive to a second excitation residual produced from the first, adaptive-codebook excitation residual. In a CELP decoder, a combined innovation codebook comprises a de-quantizer of pre-quantized coding parameters into a first excitation contribution, and a CELP innovation-codebook structure responsive to CELP innovation-codebook parameters to produce a second excitation contribution.

Claims

1. A Code-Excited Linear Prediction (CELP) codebook coding device for encoding sound into first, second, and third sets of encoding parameters, comprising:

at least one processor

a memory coupled to the processor and embodying instructions which cause the processor to implement;

a Linear Prediction (LP) filter for processing an input sound signal and producing a first LP residual;

an adaptive-codebook search module adapted to, in response to the first LP residual, find an adaptive-codebook contribution and produce pitch parameters forming the first set of encoding parameters;

a first subtractor for producing a second, adaptive-codebook excitation residual by subtracting the adaptive-code book contribution from the first LP residual;

a pre-quantizer including a calculator of a transform of the second, adaptive-codebook excitation residual to produce quantized transform coefficients forming the second set of encoding parameters;

a second subtractor for producing a third excitation residual by subtracting an inverse-transformed version of the quantized transform coefficients from the adaptive-codebook contribution; and

a CELP innovation-codebook search module adapted to, in response to the third excitation residual, find an innovation-codebook contribution and produce innovation codebook parameters forming the third set of encoding parameters.

2. A CELP codebook coding device as defined in claim 1, wherein the transform calculator of the pre-quantizer is a calculator of a Discrete Cosine Transform (DCT) of the second residual to produce DCT coefficients.

3. A CELP codebook coding device as defined in claim 1, wherein the transform calculator produces transform coefficients, and the pre-quantizer comprises a quantizer of the transform coefficients to produce the quantized transform coefficients.

4. A CELP codebook coding device as defined in claim 3, wherein the quantizer of the pre-quantizer is an algebraic vector quantizer to produce, in response to the transform coefficients from the transform calculator, algebraic-vector-quantized transform coefficients.

5. A combined innovation codebook coding device as defined in claim 3, wherein the quantizer encodes transform coefficients related to lower frequencies only, depending on an available bit-budget.

6. A CELP codebook coding device as defined in claim 1,

wherein the pre-quantizer comprises a pre-emphasis filter of the second, adaptive-codebook excitation residual to produce a pre-emphasized second residual prior to calculating the transform.

7. A CELP codebook coding device as defined in claim 6, wherein the pre-emphasis filter emphasizes low frequencies of the second, adaptive-codebook excitation residual.

8. A CELP codebook coding device as defined in claim 6, comprising a calculator of an inverse transform of the quantized transform coefficients, and a de-emphasis filter of the inverse-transformed coefficients to produce a time-domain contribution forming the inverse-transformed version of the quantized transform coefficients, and wherein the second subtractor subtracts the time-domain contribution from the adaptive-codebook contribution to produce the third residual.

9. A CELP codebook coding device as defined in claim 1, wherein the CELP innovation-codebook search module is an Algebraic CELP (ACELP) innovation-codebook search module.

10. A CELP codebook coding device as defined in claim 1, wherein the pre-quantizer quantizes only transform coefficients having an energy exceeding a specified threshold, so that spectral dynamics of the third residual are reduced or maintained within a desired range.

11. A CELP coder comprising the Code-Excited Linear Prediction (CELP) codebook coding device as defined in claim 1.

12. A Code-Excited Linear Prediction (CELP) codebook decoding device for decoding sound in response to pitch parameters, pre-quantized coding parameters and innovation-codebook parameters, comprising:

at least one processor

a memory coupled to the processor and embodying instructions which cause the processor to implement;

an adaptive codebook structure for producing an adaptive codebook contribution in response to the pitch parameters;

a de-quantizer of the pre-quantized coding parameters into a first innovation excitation contribution, the de-quantizer comprising an inverse transform calculator responsive to the coding parameters;

a CELP innovation-codebook structure responsive to the innovation-codebook parameters to produce a second innovation excitation contribution;

an adder of the adaptive codebook contribution, the first innovation excitation contribution and the second innovation excitation contribution to form a total excitation signal; and

a Linear Prediction (LP) synthesis filter for processing the total excitation signal and producing a synthesis signal representing a synthesis of said sound.

13. A CELP codebook decoding device as defined in claim 12, wherein the de-quantizer comprises a decoder for de-quantizing the pre-quantized coding parameters.

14. A CELP codebook decoding device as defined in claim 13, wherein the decoder comprises an Algebraic Vector Quantizer (AVQ) decoder.

15. A CELP codebook decoding device as defined in claim 13, wherein the inverse transform calculator is responsive to the de-quantized coding parameters.

16. A CELP codebook decoding device as defined in claim 15, wherein the inverse transform is an inverse Discrete Cosine Transform (DCT).

17. A CELP codebook decoding device as defined in claim 13, wherein the de-quantizer comprises a de-emphasis filter supplied with the inverse-transformed, de-quantized coding parameters to produce the first innovation excitation contribution.

18. A CELP decoder comprising the Code-Excited Linear Prediction (CELP) codebook decoding device according to claim 12.

19. A Code-Excited Linear Prediction (CELP) codebook coding method for coding sound into first, second and third sets of encoding parameters, comprising:

producing a first Linear Prediction (LP) residual using a sound signal;

searching an adaptive codebook in response to the first LP residual to find an adaptive-codebook contribution and produce pitch parameters forming the first set of encoding parameters;

producing a second, adaptive-codebook excitation residual by subtracting the adaptive codebook contribution from the first LP residual;

pre-quantizing the second, adaptive-codebook excitation residual, the pre-quantizing comprising calculating a transform of the second, adaptive-codebook excitation residual to produce quantized transform coefficients forming the second set of encoding parameters;

producing a third excitation residual by subtracting an inverse-transformed version of the quantized transform coefficients from the adaptive-codebook contribution; and

searching a CELP innovation-codebook in response to the third excitation residual to find an innovation-codebook contribution and produce innovation codebook parameters forming the third set of encoding parameters.

20. A CELP codebook coding method as defined in claim 19, wherein the transform is a Discrete Cosine Transform (DCT).

21. A CELP codebook coding method as defined in claim 19, wherein the pre-quantizing comprises calculating the transform of the second, adaptive-codebook excitation residual to produce transform coefficients, and quantizing the transform coefficients to produce the quantized transform coefficients.

22. A CELP codebook coding method as defined in claim 21, wherein quantizing the transform coefficients comprises algebraic vector quantizing said transform coefficients.

23. A CELP codebook coding method as defined in claim 21, wherein quantizing the transform coefficients comprises encoding transform coefficients related to lower frequencies only, depending on an available bit-budget.

24. A CELP codebook coding method as defined in claim 19, wherein pre-quantizing the second, adaptive-codebook excitation residual comprises pre-emphasis filtering the second, adaptive-codebook excitation residual prior to calculating the transform of the second, adaptive-codebook excitation residual.

25. A CELP codebook coding method as defined in claim 24, wherein pre-emphasis filtering comprises emphasizing low frequencies of the second, adaptive-codebook excitation residual.

26. A CELP codebook coding method as defined in claim 19, comprising calculating an inverse transform of the quantized transform coefficients, and de-emphasis filtering the inverse-transformed coefficients to produce a time-domain contribution forming the inverse-transformed version of the quantized transform coefficients, wherein producing the third excitation residual comprises subtracting the time-domain contribution from the adaptive-codebook contribution.

27. A CELP codebook coding method as defined in claim 19, wherein the CELP innovation-codebook search is an ACELP innovation-codebook search.

28. A CELP codebook coding method as defined in claim 19, wherein pre-quantizing the second, adaptive-codebook excitation residual comprises pre-quantizing only transform coefficients having an energy exceeding a specified threshold, so that spectral dynamics of the third excitation residual are reduced or maintained within a desired range.

29. A Code-Excited Linear Prediction (CELP) codebook decoding method for decoding sound in response to pitch parameters, pre-quantized coding parameters and innovation-codebook parameters, comprising:

applying the pitch parameters to an adaptive codebook structure for producing an adaptive codebook contribution;

de-quantizing the pre-quantized coding parameters into a first innovation excitation contribution, wherein de-quantizing the pre-quantized coding parameters comprises calculating an inverse transform of the coding parameters; and

applying the innovation-codebook parameters to a CELP innovation-codebook structure to produce a second innovation excitation contribution;

adding the adaptive codebook contribution, the first innovation excitation contribution and the second innovation excitation contribution to form a total excitation signal; and

processing the total excitation signal through a Linear Prediction (LP) synthesis filter to produce a synthesis signal representing a synthesis of said sound.

30. A CELP codebook decoding method as defined in claim 29, wherein de-quantizing the pre-quantized coding parameters comprises decoding the pre-quantized coding parameters to produce de-quantized coding parameters.

31. A CELP codebook decoding method as defined in claim 30, wherein decoding the pre-quantized coding parameters comprises Algebraic Vector Quantizer (AVQ) decoding said pre-quantized coding parameters.

32. A CELP codebook decoding method as defined in claim 29, wherein calculating an inverse transform of the coding parameters comprises calculating the inverse transform of the de-quantized coding parameters.

33. A CELP codebook decoding method as defined in claim 32, wherein the inverse transform is an inverse Discrete Cosine Transform (DCT).

34. A CELP codebook decoding method as defined in claim 32, comprising de-emphasis filtering the inverse-transformed, de-quantized coding parameters to produce the first innovation excitation contribution.