Method and apparatus to encode/decode low bit-rate audio signal by approximiating high frequency envelope with strongly correlated low frequency codevectors

• US 8,301,439 B2
• Filed: 07/12/2006
• Issued: 10/30/2012
• Est. Priority Date: 07/15/2005
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Kim, Junghoe, Oh, Eunmi, Kudryashov, Boris, Osipov, Konstantin
• Primary Examiner(s): Smits, Talivaldis Ivars
• Assistant Examiner(s): Kazeminezhad, Farzad
• Application Number:
  US11/456,946
• Publication Number:
  US 20070016411A1
• Time in Patent Office: 2,302 Days
• Field of Search: 704/226, 704/265
• US Class Current: 704/226
• CPC Class Codes: G10L 19/0017 Lossless audio signal coding; Perfect reconstruction of coded audio signal by transmission of coding error G10L19/24 takes precedence
• CPC Class Codes: G10L 19/02 using spectral analysis, e.g. transform vocoders or subband vocoders
• CPC Class Codes: G10L 19/06 Determination or coding of the spectral characteristics, e.g. of the short-term prediction coefficients

Abstract

A method of encoding a low bit-rate audio signal includes quantizing and encoding a plurality of low frequency sub-bands of an audio signal in a frequency domain, generating a codebook of codevectors using sub-bands of the audio signal spectrum, detecting an envelope of another frequency sub-band of the audio signal and quantizing and losslessly-encoding the detected envelope, selecting a codevector most similar to the higher frequency sub-band spectrum from the generated codebook'"'"'s codevectors and determining its codebook codevector index, and generating a bit stream. Decoding the low bit-rate audio signal includes restoring and dividing a bit stream into a plurality of first frequency sub-bands and at least one second frequency sub-band and inversely quantizing the first frequency sub-bands in the bit stream, restoring codebook codevector index information and envelope information for the second frequency sub-band, generating a codebook of codevectors using the inversely quantized first frequency sub-bands, and restoring the second frequency sub-band using the restored codevector index information and the envelope information.

Claims

1. A method of encoding a low bit-rate audio signal, the method comprising:

quantizing and encoding a plurality of first frequency sub-bands in a low frequency region of a full-band audio signal having a plurality of frequency sub-bands;

generating a codebook of codevectors using one or more of the plurality of overlapping first frequency sub-bands of the low frequency region of the full-band audio signal to vector quantize at least one high frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution;

detecting an envelope of at least one second frequency sub-band in a high frequency region of the full-band audio signal higher than the highest low frequency sub-band and quantizing and encoding the envelope;

selecting a codevector in the first frequency sub-band codebook that is most similar to the second frequency sub-band to be encoded from the generated codebook'"'"'s codevectors and determining a codebook codevector index which defines a fine structure of the first frequency sub-band to help encode the structure of the similar second frequency sub-band; and

generating a bit stream using encoded data generated in the encoding of the first frequency sub-band, the envelope, and the determined codebook codevector index.

2. The method of claim 1, wherein the quantizing and encoding of the first frequency sub-band is one of an MPEG1 layer 3 coding (mp3) and an MPEG-2/4 advanced audio coding (AAC).

3. A method of encoding a low bit-rate audio signal, the method comprising:

quantizing and encoding a plurality of first frequency sub-bands in a low frequency region of a full-band audio signal having a plurality of frequency sub-bands;

generating a codebook of codevectors using one or more of the plurality of first frequency sub-bands of the low frequency region of the full band audio signal to vector quantize at least one high-frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution;

detecting an envelope of a second frequency sub-band in the high frequency region of the full-band audio signal higher than the highest low frequency sub-band and quantizing and encoding the detected envelope of the second frequency sub-band;

checking whether a codevector having at least a predetermined similarity exists among the generated codevectors of the codebook with respect to a high frequency band to be encoded;

if the similar codevector exists, selecting the similar codevector, determining a codebook codevector index which defines a fine structure of the first frequency sub-band to help encode the structure of the similar second frequency sub-band;

if a similar codevector does not exist, encoding information indicating that a similar codevector does not exist; and

generating a bit stream using encoded data generated in the encoding of the first frequency sub-band, the envelope of the second frequency sub-band, the determined codebook codevector index, and the information indicating that the similar codebook does not exist.

4. The method of claim 3, wherein the first frequency sub-band is a low frequency sub-band.

5. The method of claim 3, wherein the high frequency band is a non-uniform band which takes hearing characteristics into consideration.

6. The method of claim 5, wherein the non-uniform band is a bark band.

7. The method of claim 3, wherein the codebooks are generated using overlapped spectra.

8. The method of claim 3, wherein the similarity is determined using a Euclidian distance or a correlation between the codebook codevectors.

9. The method of claim 3, further comprising:

generating the audio signal in a frequency domain by converting an audio signal in a time domain to the audio signal in the frequency domain.

10. An apparatus to encode a low bit-rate audio signal, the apparatus comprising:

a low frequency quantizing/encoding unit which quantizes and encodes a plurality of first frequency sub-bands in a low frequency region of a full-band audio signal having a plurality of frequency sub-bands;

a codebook generating unit which generates a codebook of codevectors using one or more of the plurality of first frequency sub-bands of the low frequency region of the full-band audio signal to vector quantize at least one high frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution;

an envelope quantizing/encoding unit which detects an envelope of at least one second frequency sub-band in a high frequency region of the full-band audio signal higher than the highest low frequency sub-band and quantizes and encodes the detected envelope of the second frequency sub-band;

a codebook index acquiring unit which selects a codevector in the first frequency sub-band codebook most similar to the second frequency sub-band to be encoded from among the generated codebook'"'"'s codevectors and determines a codebook codevector index to define a fine structure of the first frequency sub-band to help encode the structure of the similar second frequency sub-band; and

a bit stream generating unit which generates a bit stream using encoded data which are generated by the low frequency quantizing/encoding unit, and the envelope quantizing/encoding unit and the determined codebook codevector index.

11. The apparatus of claim 10, wherein the low frequency quantizing/encoding unit quantizes and encodes using one of an MPEG1 layer 3 coding (mp3) and an MPEG-2/4 advanced audio coding (AAC).

12. An apparatus to encode a low bit-rate audio signal, the apparatus comprising:

a low frequency quantizing/encoding unit which quantizes and encodes a plurality of first frequency sub-bands in a low frequency region of a full-band audio signal having a plurality of frequency sub-bands;

a codebook generating unit which generates a codebook of codevectors using one or more of the plurality of first frequency sub-bands of the low frequency region of the full-band audio signal to vector quantize at least one high frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution;

an envelope quantizing/encoding unit which detects an envelope of at least one second frequency sub-band of a high frequency region of the full-band audio signal higher than the highest low frequency sub-band and quantizes and encodes the detected envelope of the second frequency sub-band;

a similarity checking unit which checks whether a codevector of the codebook having at least a predetermined similarity exists among the generated codevectors of the codebook with respect to a high frequency band to be encoded;

a codebook existence information/index encoding unit which selects a similar codevector from the codebook, determines a codebook codevector index which defines a fine structure of the first frequency sub-band if the similar codevector in the codebook exists, and encodes the determined codebook codevector index and information indicating that the similar codevector exists;

a codebook existence information encoding unit which encodes information indicating that a similar codevector in the codebook does not exist if a similar codevector in the codebook does not exist; and

a bitstream generating unit which generates a bit stream using encoded data which are generated by the low frequency quantizing/encoding unit, the codebook existence information/index encoding unit, and the codebook existence information encoding unit.

13. The apparatus of claim 12, wherein the first frequency sub-band is a low frequency sub-band.

14. The apparatus of claim 12, wherein the first frequency sub-band is a non-uniform band which takes hearing characteristics into consideration.

15. The apparatus of claim 14, wherein the non-uniform band is a bark band.

16. The apparatus of claim 12, wherein the codebooks are generated using overlapped spectra.

17. The apparatus of claim 12, wherein the similarity is determined using a Euclidian distance or a correlation between the codebooks.

18. The apparatus of claim 12, further comprising:

a T/F converting unit which converts an audio signal in a time domain into the audio signal in a frequency domain.

19. An encoding apparatus, comprising:

a first quantizing/encoding unit to quantize a plurality of first frequency sub-bands of a full spectrum audio signal having a plurality of frequency sub-bands and to encode the quantized first frequency sub-bands;

a second quantizing/encoding unit to quantize one or more envelopes of at least one second frequency sub-band of the full spectrum audio signal and to encode the quantized one or more envelopes;

a codebook unit to generate a codebook of one or more codevectors from the first frequency sub-bands to vector quantize the at least one second frequency sub-band of the plurality of frequency sub-bands or to encode the at least one second frequency sub-band by perceptual noise substitution, to determine whether a similar codevector in the codebook exists for the second frequency sub-band, and to encode codevector existence information to indicate similarities between the second frequency sub-bands and the codevectors in the codebook; and

a bit stream unit to generate a bitstream including the encoded first frequency sub-band, the encoded envelopes of the at least one second frequency sub-band, and the encoded codevector existence information.

20. The encoding apparatus of claim 19, wherein the second quantizing/encoding unit encodes the envelopes of the second frequency sub-band using the vector quantization when the corresponding second bands in the second frequency sub-band are determined to be similar to ones of the codevectors, and encodes the envelopes of the second frequency sub-band using the perceptual noise substitution when the corresponding second bands in the second frequency sub-band are determined not to be similar to any of the codebooks.

21. A method of decoding a low bit-rate audio signal, the method comprising:

dividing a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands, and at least one second frequency sub-band other than the first frequency sub-band of the full-band audio signal;

decoding and inversely quantizing the first frequency sub-bands of the full-band audio signal;

restoring codebook codevector index information which defines a fine structure of the first frequency sub-band to help encode the structure of the similar second frequency sub-band and envelope information about the second frequency sub-band;

generating a codebook of codevectors using the inversely quantized first frequency sub-band to inversely vector quantize the at least one second frequency sub-band or to decode the at least one second frequency sub-band by perceptual noise substitution; and

restoring the second frequency sub-band using the restored codebook codevector index information and the restored envelope information about the second frequency sub-band.

22. The method of claim 21, wherein the decoding and inverse quantizing of the first frequency sub-band is one of an MPEG1 layer 3 decoding (mp3) and an MPEG-2/4 advanced audio decoding (AAC).

23. A method of decoding a low bit-rate audio signal, the method comprising:

dividing a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands and at least one second frequency sub-band other than the first frequency sub-band of the full-band audio signal;

decoding and inversely quantizing the first frequency sub-bands of the full-band audio signal;

decoding information as to whether a similar codevector in a codebook exists;

if a similar codevector in the codebook exists, restoring codebook codevector index information that defines a fine structure of the first frequency sub-band and envelope information about the second frequency sub-band to help decode the structure of the similar second frequency sub-band;

generating a codebook of codevectors to inversely vector quantize at least one second frequency sub-band or to decode the at least one second frequency sub-band by perceptual noise substitution using the first frequency sub-bands which are decoded and inversely quantized and restoring the at least one second frequency sub-band using the restored codebook index information that defines a fine structure of the first frequency sub-band and the restored envelope information about the second frequency sub-band; and

if a similar codebook does not exist, restoring the envelope information and restoring the second frequency sub-band using a signal of a previous band and the restored envelope information.

24. The method of claim 23, wherein the first frequency sub-band is a low frequency sub-band.

25. The method of claim 23, wherein the previous band is a non-uniform band which takes hearing characteristics into consideration.

26. The method of claim 25, wherein the non-uniform band is a bark band.

27. The method of claim 23, wherein the codebooks are generated using overlapped spectra.

28. The method of claim 23, wherein the similarity is determined using a Euclidian distance or a correlation between the codebooks.

29. The method of claim 23, further comprising:

generating the audio signal by inversely converting an audio signal in a frequency domain into an audio signal in a time domain.

30. An apparatus to decode a low bit-rate audio signal, the apparatus comprising:

a bit stream dividing unit which divides a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands and at least one second frequency sub-band other than the first frequency sub-bands of the full-band audio signal;

a low frequency restoring unit which decodes and inversely quantizes the first frequency sub-bands of the full-band audio signal;

a high frequency index/envelope restoring unit which restores codebook codevector index information which defines a fine structure of the first frequency sub-band and envelope information about the second frequency sub-band to help restore the structure of the similar second frequency sub-band;

a codebook generating unit which generates a codebook of codevectors using the first frequency sub-band inversely quantized in the low frequency restoring unit to inversely vector quantize at least one second frequency sub-band or to decode at least one second frequency sub-band by perceptual noise substitution; and

a high frequency restoring unit which restores the second frequency sub-band using the restored codebook codevector index information that defines a fine structure of the second frequency sub-band and the restored envelope information about the other frequency sub-band.

31. The apparatus of claim 30, wherein the quantizing and decoding and inverse quantization is one of an MPEG1 layer 3 decoding (mp3) and an MPEG-2/4 advanced audio decoding (AAC).

32. An apparatus to decode a low bit-rate audio signal, the apparatus comprising:

a bit stream dividing unit which divides a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands and at least one second frequency sub-band other than the first frequency sub-bands of the full-band audio signal;

a low frequency restoring unit which decodes and inversely quantizes the first frequency sub-bands of the full-band audio signal;

a codebook existence information restoring unit which decodes information as to whether a similar codevector of a codebook exists to inversely vector quantize at least one second frequency sub-band or to decode at least one second frequency sub-band by perceptual noise substitution;

an index/envelope restoring unit which, if the similar codevector of the codebook exists, restores codebook codevector index information that defines a fine structure of the first frequency sub-band and envelope information about the second frequency sub-band to restore the similar second frequency sub-band;

a first high frequency restoring unit which generates a codebook of codevectors using the first frequency sub-band which is decoded and inversely quantizes and restores the second frequency sub-band using the restored codebook index information and the restored envelope information about the second frequency sub-band; and

a second high frequency restoring unit which, if a similar codevector does not exist, restores the envelope information and restores the second frequency sub-band using a signal of a previous band and the restored envelope information.

33. The apparatus of claim 32, wherein the first frequency sub-band is a low frequency sub-band.

34. The apparatus of claim 32, wherein the previous band is a non-uniform band which takes hearing characteristics into consideration.

35. The apparatus of claim 34, wherein the non-uniform band is a bark band.

36. The apparatus of claim 32, wherein the codebooks are generated using overlapped spectra.

37. The apparatus of claim 32, wherein the similarity is determined using a Euclidian distance or a correlation between the codebooks.

38. The apparatus of claim 32, further comprising:

an F/T converting unit which inversely converts the audio signal from an audio signal in a frequency domain into an audio signal in a time domain.

39. A non-transitory computer-readable medium having a computer executable program for a method of encoding a low bit-rate audio signal, the method comprising:

quantizing and encoding a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands in a low frequency region of the full-band audio signal;

generating a codebook of codevectors using the first frequency sub-bands in the low frequency region to vector quantize at least one high frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution;

detecting an envelope of at least one second frequency sub-band of the full-band audio signal other than the first frequency sub-bands in a high frequency region and quantizing and encoding the envelope;

selecting a codevector in the first frequency sub-band codebook that is most similar to the second frequency sub-band to be encoded from the codebook'"'"'s codevectors and determining a codebook codevector index which defines a fine structure of the first frequency sub-band to help encode the structure of the similar second frequency sub-band; and

generating a bit stream using encoded data generated in the encoding of the first frequency sub-band, the envelope, and the determined codebook codevector index.

40. A non-transitory computer-readable medium having a computer executable program for a method of encoding a low bit-rate audio signal, the method comprising:

quantizing and encoding a plurality of first frequency sub-bands in a low frequency region of the full-band audio signal having a plurality of frequency sub-bands;

generating a codebook of codevectors using the first frequency sub-bands in the low frequency region to vector quantize at least one high frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution;

detecting an envelope of at least one second frequency sub-bands of the full-band audio signal other than the first frequency sub-bands in a high-frequency region and quantizing and encoding the detected envelope of the second frequency sub-band;

checking whether a codevector of a codebook having at least a predetermined similarity exists among the generated codevectors of the codebook with respect to the first frequency sub-band to help a similar second frequency sub-band to be encoded;

if the similar codevector of the codebook exists, selecting the similar codevector of the codebook, determining a codebook codevector index which defines a fine structure of the first frequency sub-band to help encode the similar second frequency sub-band, and encoding the determined codebook index and information indicating that the similar codebook exists;

if a similar codevector of the codebook does not exist, encoding information indicating that a similar codevector of the codebook does not exist; and

generating a bit stream using encoded data generated in the encoding of the first frequency sub-band, the envelope of the second frequency sub-band, the determined codebook codevector index, and the information indicating that the similar codebook codevector does not exist.

41. A non-transitory computer-readable medium having a computer executable program for a method of decoding a low bit-rate audio signal, the method comprising:

dividing a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands and at least one second frequency sub-band higher than the highest first frequency sub-band of the full-band audio signal;

decoding and inversely quantizing the first frequency sub-band of the full-band audio signal;

restoring codebook codevector index information that defines a fine structure of the first frequency sub-band and envelope information about the second frequency sub-band to help restore the structure of the similar second frequency sub-band;

generating a codebook of codevectors to inversely vector quantize at least one second frequency sub-band or to decode at least one second frequency sub-band by perceptual noise substitution using the inversely quantized first frequency sub-band; and

restoring the second frequency sub-band using the restored codebook codevector index information and the restored envelope information about the second frequency sub-band.

42. A non-transitory computer-readable medium having a computer executable program for a method of decoding a low bit-rate audio signal, the method comprising:

dividing a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands and at least one second frequency sub-band higher than the highest first frequency sub-band of the full-band audio signal;

decoding and inversely quantizing the first frequency sub-bands of the full-band audio signal;

decoding information as to whether a similar codevector of a codebook exists;

if a similar codevector of the codebook exists, restoring codebook codevector index information that defines the fine structure of the first frequency sub-band and envelope information about the second frequency sub-band to help restore the structure of the similar second frequency sub-band;

generating a codebook of codevectors to inversely vector quantize at least one second frequency sub-band or to decode at least one second frequency sub-band by perceptual noise substitution using the first frequency sub-bands which are decoded and inversely quantized and restoring the second frequency sub-band using the restored codebook codevector index information and the restored envelope information about the second frequency sub-band; and

if a similar codevector in the codebook does not exist, restoring the envelope information and restoring the second frequency sub-band using a signal of a previous band and the restored envelope information.

43. A method of encoding a full-band audio signal, the method comprising:

encoding a plurality of first frequency sub-bands of the full-band audio signal having a plurality of frequency sub-bands;

generating a codebook of codevectors to vector quantize at least one high frequency sub-band or to encode at least one high frequency sub-band by perceptual noise substitution using the first frequency sub-bands;

determining an envelope of at least one second frequency sub-band of the full-band audio signal which is different than the first frequency sub-band and encoding the envelope;

selecting one of the generated codevectors of the codebook if it has a predetermined similarity to the second frequency sub-band; and

encoding a bit stream with the encoded first frequency sub-bands, the encoded envelope, and one of an indication that a codevector of the codebook was selected with an index of the selected codebook which defines a fine structure of the first frequency sub-band to help encode the structure of the similar second frequency sub-band and an indication that a codebook was not selected.

44. A method of decoding an audio signal, the method comprising:

dividing a bit stream into a plurality of first frequency sub-bands of a full-band audio signal having a plurality of frequency sub-bands and at least one second frequency sub-band having envelope information of the full-band audio signal;

decoding the first frequency sub-band and generating codevectors of a codebook to inversely quantize at least one high frequency sub-band or to decode at least one high frequency sub-band by perceptual noise substitution using the decoded first frequency sub-band;

decoding codebook information which defines a fine structure of the first frequency sub-band indicating whether a codevector of the generated codebook has a similarity to the first frequency sub-band to help decode the structure of the similar second frequency sub-band;

restoring the second frequency sub-band based on the codebook information using one of a generated codevectors of the codebook with the envelope information and a signal of a previous band with the envelope information.