High-band signal modeling

• US 10,163,447 B2
• Filed: 12/12/2014
• Issued: 12/25/2018
• Est. Priority Date: 12/16/2013
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Krishnan, Venkatesh, Atti, Venkatraman S.
• Primary Examiner(s): Neway, Samuel G
• Application Number:
  US14/568,359
• Publication Number:
  US 20150170662A1
• Time in Patent Office: 1,474 Days
• CPC Class Codes: G10L 19/02 using spectral analysis, e.g. transform vocoders or subband vocoders
• CPC Class Codes: G10L 19/0204 using subband decomposition
• CPC Class Codes: G10L 19/04 using predictive techniques
• CPC Class Codes: G10L 19/08 Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters
• CPC Class Codes: G10L 21/038 using band spreading techniques
• CPC Class Codes: G10L 21/0388 Details of processing therefor

Abstract

A method includes filtering, at a speech encoder, an audio signal into a first group of sub-bands within a first frequency range and a second group of sub-bands within a second frequency range. The method also includes generating a harmonically extended signal based on the first group of sub-bands. The method further includes generating a third group of sub-bands based, at least in part, on the harmonically extended signal. The third group of sub-bands corresponds to the second group of sub-bands. The method also includes determining a first adjustment parameter for a first sub-band in the third group of sub-bands or a second adjustment parameter for a second sub-band in the third group of sub-bands. The first adjustment parameter is based on a metric of a first sub-band in the second group of sub-bands, and the second adjustment parameter is based on a metric of a second sub-band in the second group of sub-bands.

Claims

1. A method of reducing a transmission bandwidth of a bit stream, the method comprising:

filtering, at a speech encoder, an audio signal into a group of low-frequency sub-bands within a low-band frequency range and a first group of high-frequency sub-bands within a high-band frequency range;

generating a first residual signal of a first high-frequency sub-band in the first group of high-frequency sub-bands;

generating a harmonically extended signal based on the group of low-frequency sub-bands and a non-linear processing function;

generating a second group of high-frequency sub-bands based, at least in part, on the harmonically extended signal, wherein the second group of high-frequency sub-bands corresponds to the first group of high-frequency sub-bands;

determining, at a dedicated parameter estimator, a first adjustment parameter based on a comparison of an energy level associated with the first residual signal to an energy level of a first high-frequency sub-band in the second group of high-frequency sub-bands;

determining a second adjustment parameter for a second high-frequency sub-band in the second group of high-frequency sub-bands based on a metric of a second high-frequency sub-band in the first group of high-frequency sub-bands; and

transmitting the first adjustment parameter and the second adjustment parameter to a speech decoder as part of the bit stream, the first adjustment parameter and the second adjustment parameter usable by the speech decoder to reconstruct the first group of high-frequency sub-bands, wherein the transmission bandwidth of the bit stream is reduced compared to transmission of an encoded version of the first group of high-frequency sub-bands.

2. The method of claim 1, wherein the first adjustment parameter and the second adjustment parameter correspond to gain adjustment parameters.

3. The method of claim 1, wherein the first adjustment parameter and the second adjustment parameter correspond to linear prediction coefficient adjustment parameters.

4. The method of claim 1, wherein the first adjustment parameter and the second adjustment parameter correspond to time varying envelope adjustment parameters.

5. The method of claim 1, further comprising inserting the first adjustment parameter and the second adjustment parameter into an encoded version of the audio signal to enable adjustment during reconstruction of the audio signal from the encoded version of the audio signal.

6. The method of claim 1, wherein generating the second group of high-frequency sub-bands comprises:

mixing the harmonically extended signal with modulated noise to generate a high-band excitation signal, wherein the modulated noise and the harmonically extended signal are mixed based on a mixing factor; and

filtering the high-band excitation signal into the second group of high-frequency sub-bands.

7. The method of claim 6, wherein the mixing factor is determined based on at least one among a pitch lag, an adaptive codebook gain associated with the group of low-frequency sub-bands, a pitch correlation between the group of low-frequency sub-bands and the first group of high-frequency sub-bands.

8. The method of claim 1, wherein generating the second group of high-frequency sub-bands comprises:

filtering the harmonically extended signal into a plurality of sub-bands; and

mixing each sub-band of the plurality of sub-bands with modulated noise to generate a plurality of high-band excitation signals, wherein the plurality of high-band excitation signals corresponds to the second group of high-frequency sub-bands.

9. The method of claim 8, wherein the modulated noise and a first sub-band of the plurality of sub-bands are mixed based on a first mixing factor, and wherein the modulated noise and a second sub-band of the plurality of sub-bands are mixed based on a second mixing factor.

10. An apparatus for reducing a transmission bandwidth of a bit stream, the apparatus comprising:

a first filter configured to filter an audio signal into a group of low-frequency sub-bands within a low-band frequency range and a first group of high-frequency sub-bands within a high-band frequency range;

a parameter estimator configured to generate a first residual signal of a first high-frequency sub-band in the first group of high-frequency sub-bands;

a non-linear transformation generator configured to generate a harmonically extended signal based on the group of low-frequency sub-bands and a non-linear processing function;

a second filter configured to generate a second group of high-frequency sub-bands based, at least in part, on the harmonically extended signal, wherein the second group of high-frequency sub-bands corresponds to the first group of high-frequency sub-bands;

dedicated parameter estimators configured to;

determine a first adjustment parameter based on a comparison of an energy level associated with the first residual signal to an energy level of a first high-frequency sub-band in the second group of high-frequency sub-bands; and

determine a second adjustment parameter for a second high-frequency sub-band in the second group of high-frequency sub-bands based on a metric of a second high-frequency sub-band in the first group of high-frequency sub-bands; and

a transmitter to transmit the first adjustment parameter and the second adjustment parameter to a speech decoder as part of the bit stream, the first adjustment parameter and the second adjustment parameter usable by the speech decoder to reconstruct the first group of high-frequency sub-bands, wherein the transmission bandwidth of the bit stream is reduced compared to transmission of an encoded version of the first group of high-frequency sub-bands.

11. The apparatus of claim 10, wherein the first adjustment parameter and the second adjustment parameter correspond to gain adjustment parameters.

12. The apparatus of claim 10, wherein the first adjustment parameter and the second adjustment parameter correspond to linear prediction coefficient adjustment parameters.

13. The apparatus of claim 10, wherein the first adjustment parameter and the second adjustment parameter correspond to time varying envelope adjustment parameters.

14. The apparatus of claim 10, further comprising a multiplexer configured to insert the first adjustment parameter and the second adjustment parameter into an encoded version of the audio signal to enable adjustment during reconstruction of the audio signal from the encoded version of the audio signal.

15. The apparatus of claim 10, wherein generating the second group of high-frequency sub-bands comprises:

mixing the harmonically extended signal with modulated noise to generate a high-band excitation signal, wherein the modulated noise and the harmonically extended signal are mixed based on a mixing factor; and

filtering the high-band excitation signal into the second group of high-frequency sub-bands.

16. The apparatus of claim 15, wherein the mixing factor is determined based on at least one among a pitch lag, an adaptive codebook gain associated with the group of low-frequency sub-bands, and a pitch correlation between the group of low-frequency sub-bands and the first group of high-frequency sub-bands.

17. The apparatus of claim 10, wherein generating the second group of high-frequency sub-bands comprises:

filtering the harmonically extended signal into a plurality of sub-bands; and

mixing each sub-band of the plurality of sub-bands with modulated noise to generate a plurality of high-band excitation signals, wherein the plurality of high-band excitation signals corresponds to the second group of high-frequency sub-bands.

18. The apparatus of claim 17, wherein the modulated noise and a first sub-band of the plurality of sub-bands are mixed based on a first mixing factor, and wherein the modulated noise and a second sub-band of the plurality of sub-bands are mixed based on a second mixing factor.

19. A non-transitory computer-readable medium comprising instructions for reducing a transmission bandwidth of a bit stream, wherein the instructions, when executed by a processor at a speech encoder, cause the processor to:

filter an audio signal into a group of low-frequency sub-bands within a low-band frequency range and a first group of high-frequency sub-bands within a high-band frequency range;

generate a first residual signal of a first sub-band in the first group of high-frequency sub-bands;

generate a harmonically extended signal based on the group of low-frequency sub-bands and a non-linear processing function;

generate a second group of high-frequency sub-bands based, at least in part, on the harmonically extended signal, wherein the second group of high-frequency sub-bands corresponds to the first group of high-frequency sub-bands;

determine, at a dedicated parameter estimator, a first adjustment parameter based on a comparison of an energy level associated with the first residual signal to an energy level of a first high-frequency sub-band in the second group of high-frequency sub-bands;

determine a second adjustment parameter for a second high-frequency sub-band in the second group of high-frequency sub-bands based on a metric of a second high-frequency sub-band in the first group of high-frequency sub-bands; and

initiate transmission of the first adjustment parameter and the second adjustment parameter to a speech decoder as part of the bit stream, wherein the first adjustment parameter and the second adjustment parameter are usable by the speech decoder to reconstruct the first group of high-frequency sub-bands, and wherein the transmission bandwidth of the bit stream is reduced compared to transmission of an encoded version of the first group of high-frequency sub-bands.

20. The non-transitory computer-readable medium of claim 19, wherein the first adjustment parameter and the second adjustment parameter correspond to gain adjustment parameters.

21. The non-transitory computer-readable medium of claim 19, wherein the first adjustment parameter and the second adjustment parameter correspond to linear prediction coefficient adjustment parameters.

22. The non-transitory computer-readable medium of claim 19, wherein the first adjustment parameter and the second adjustment parameter correspond to time varying envelope adjustment parameters.

23. The non-transitory computer-readable medium of claim 19, further comprising instructions that, when executed by the processor, cause the processor to insert the first adjustment parameter and the second adjustment parameter into an encoded version of the audio signal to enable adjustment during reconstruction of the audio signal from the encoded version of the audio signal.

24. An apparatus for reducing a transmission bandwidth of a bit stream, the apparatus comprising:

means for filtering an audio signal into a group of low-frequency sub-bands within a low-band frequency range and a first group of high-frequency sub-bands within a high-band frequency range;

means for generating a first residual signal of a first high-frequency sub-band in the first group of high-frequency sub-bands;

means for generating a harmonically extended signal based on the group of low-frequency sub-bands and a non-linear processing function;

means for generating a second group of high-frequency sub-bands based, at least in part, on the harmonically extended signal, wherein the second group of high-frequency sub-bands corresponds to the first group of high-frequency sub-bands;

means for determining a first adjustment parameter based on a comparison of an energy level associated with the first residual signal to an energy level of a first high-frequency sub-band in the second group of high-frequency sub-bands;

means for determining a second adjustment parameter for a second high-frequency sub-band in the second group of high-frequency sub-bands based on a metric of a second high-frequency sub-band in the first group of high-frequency sub-bands; and

means for transmitting the first adjustment parameter and the second adjustment parameter to a speech decoder as part of the bit stream, the first adjustment parameter and the second adjustment parameter usable by the speech decoder to reconstruct the first group of high-frequency sub-bands, wherein the transmission bandwidth of the bit stream is reduced compared to transmission of an encoded version of the first group of high-frequency sub-bands.

25. The apparatus of claim 24, wherein the first adjustment parameter and the second adjustment parameter correspond to gain adjustment parameters.

26. The apparatus of claim 24, wherein the first adjustment parameter and the second adjustment parameter correspond to linear prediction coefficient adjustment parameters.

27. The apparatus of claim 24, wherein the first adjustment parameter and the second adjustment parameter correspond to time varying envelope adjustment parameters.

28. The apparatus of claim 24, further comprising means for inserting the first adjustment parameter and the second adjustment parameter into an encoded version of the audio signal to enable adjustment during reconstruction of the audio signal from the encoded version of the audio signal.

29. A method comprising:

generating, at a speech decoder, a harmonically extended signal based on a low-band excitation signal, wherein the low-band excitation signal is generated by a linear prediction based decoder based on parameters received from a speech encoder;

generating a group of high-band excitation sub-bands based, at least in part, on the harmonically extended signal;

adjusting, at a dedicated parameter adjuster, the group of high-band excitation sub-bands based on adjustment parameters received from the speech encoder, wherein a transmission bandwidth of a bit stream is reduced compared to transmission of an encoded version of high-frequency sub-bands of an encoder-side audio signal, and wherein the adjustment parameters comprise;

a first adjustment parameter based on a comparison of an energy level of a first high-frequency sub-band in a group of high-frequency sub-bands to an energy level associated with a residual signal of a first high-frequency sub-band in a second group of high-frequency; and

a second adjustment parameter for a second high-frequency sub-band in the group of high-frequency sub-bands; and

reconstructing the high-frequency sub-bands of the encoder-side audio signal based on the adjusted group of high-band excitation sub-bands.

30. The method of claim 29, wherein the adjustment parameters include gain adjustment parameters, linear predication coefficient adjustment parameters, time varying envelope adjustment parameters, or a combination thereof.

31. An apparatus comprising:

a non-linear transformation generator configured to generate a harmonically extended signal based on a low-band excitation signal, wherein the low-band excitation signal is generated by a linear prediction based decoder based on parameters received from a speech encoder;

a second filter configured to generate a group of high-band excitation sub-bands based, at least in part, on the harmonically extended signal;

dedicated parameter adjusters configured to adjust the group of high-band excitation sub-bands based on adjustment parameters received from the speech encoder, wherein a transmission bandwidth of a bit stream is reduced compared to transmission of an encoded version of high-frequency sub-bands of an encoder-side audio signal, and wherein the adjustment parameters comprise;

a first adjustment parameter based on a comparison of an energy level of a first high-frequency sub-band in a group of high-frequency sub-bands to an energy level associated with a residual signal of a first high-frequency sub-band in a second group of high-frequency; and

a second adjustment parameter for a second high-frequency sub-band in the group of high-frequency sub-bands; and

a reconstruction unit configured to reconstruct the high-frequency sub-bands of the encoder-side audio signal based on the adjusted group of high-band excitation sub-bands.

32. The apparatus of claim 31, wherein the adjustment parameters include gain adjustment parameters, linear predication coefficient adjustment parameters, time varying envelope adjustment parameters, or a combination thereof.

33. An apparatus comprising:

means for generating a harmonically extended signal based on a low-band excitation signal, wherein the low-band excitation signal is generated by a linear prediction based decoder based on parameters received from a speech encoder;

means for generating a group of high-band excitation sub-bands based, at least in part, on the harmonically extended signal;

means for adjusting the group of high-band excitation sub-bands based on adjustment parameters received from the speech encoder, wherein a transmission bandwidth of a bit stream is reduced compared to transmission of an encoded version of high-frequency sub-bands of an encoder-side audio signal, and wherein the adjustment parameters comprise;

a first adjustment parameter based on a comparison of an energy level of a first high-frequency sub-band in a group of high-frequency sub-bands to an energy level associated with a residual signal of a first high-frequency sub-band in a second group of high-frequency; and

a second adjustment parameter for a second high-frequency sub-band in the group of high-frequency sub-bands; and

means for reconstructing the high-frequency sub-bands of the encoder-side audio signal based on the adjusted group of high-band excitation sub-bands.

34. The apparatus of claim 33, wherein the adjustment parameters include gain adjustment parameters, linear predication coefficient adjustment parameters, time varying envelope adjustment parameters, or a combination thereof.

35. A non-transitory computer-readable medium comprising instructions that, when executed by a processor at a speech decoder, cause the processor to:

generate a harmonically extended signal based on a low-band excitation signal, wherein the low-band excitation signal is generated by a linear prediction based decoder based on parameters received from a speech encoder;

generate a group of high-band excitation sub-bands based, at least in part, on the harmonically extended signal; and

adjust, at a dedicated parameter adjuster, the group of high-band excitation sub-bands based on adjustment parameters received from the speech encoder, wherein a transmission bandwidth of a bit stream is reduced compared to transmission of an encoded version of high-frequency sub-bands of an encoder-side audio signal, and wherein the adjustment parameters comprise;

a first adjustment parameter based on a comparison of an energy level of a first high-frequency sub-band in a group of high-frequency sub-bands to an energy level associated with a residual signal of a first high-frequency sub-band in a second group of high-frequency; and

a second adjustment parameter for a second high-frequency sub-band in the group of high-frequency sub-bands; and

reconstruct the high-frequency sub-bands of the encoder-side audio signal based on the adjusted group of high-band excitation sub-bands.