Adaptive subband scaling method and apparatus for quantization bit allocation in variable length perceptual coding

• US 5,930,750 A
• Filed: 01/23/1997
• Issued: 07/27/1999
• Est. Priority Date: 01/30/1996
• Status: Expired due to Term

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Tsutsui, Kyoya
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/786,326
• Time in Patent Office: 915 Days
• Field of Search: 704/229, 704/230, 704/500
• US Class Current: 704/229
• CPC Class Codes: G06T 9/005 Statistical coding, e.g. Huffman, run length coding
• CPC Class Codes: G06T 9/007 Transform coding, e.g. discrete cosine transform
• CPC Class Codes: H04B 1/665 using psychoacoustic properties of the ear, e.g. masking effect

Abstract

A signal encoding method for encoding input digital data by so-called high-efficiency encoding, wherein it is possible to suppress degradation in the sound quality in case of raising the compression ratio. The range of values of subband-based signal components for quantization to zero, to which a short codelength is given, is set so as to be larger. Consequently the ratio of the subband-based signal components quantized to values of a shorter codelength will become higher than if the subband-based signal components are quantized in such a manner as to minimize the total quantization error energy in each frequency band.

Claims

1. A signal encoding method comprising the steps of:

transforming an input signal to a frequency spectrum of signal components;

splitting the signal components of the frequency spectrum into a plurality of frequency subbands;

quantizing the signal components in each subband to generate quantized values;

determining the quantization precision needed for the signal components of each subband, wherein the quantization precision is determined to be higher where the quantized values are larger and lower where the quantized values are smaller; and

encoding the quantized values in variable-length codes, wherein the code lengths are set to be longer if the quantization precision value is higher, and shorter if the quantization precision value is lower.

2. The signal encoding method of claim 1, further comprising the steps of:

calculating a total number of bits necessary for minimizing the total quantization error energy in each subband;

comparing the total number of bits necessary for minimizing the total quantization energy in each subband to a total number of available bits; and

setting a range of subband-based signal components which have a short codelength to be larger than if the subband-based signal components are quantized in such a manner as to reduce the total quantization error energy in each frequency subband only if the total number of bits necessary for minimizing the total quantization energy in each subband is larger than the total number of available bits.

3. The signal encoding method of claim 1 further comprising the steps of:

selecting a normalizing coefficient; and

prior to the step of quantizing, normalizing the signal components of each subband.

4. The signal encoding method of claim 3, wherein the normalization coefficient of the normalizing step approximates the maximum absolute value of the spectral signal component of each subband.

5. The signal encoding method of claim 3, wherein the steps of normalizing, quantizing, determining quantization precision and encoding begin in a highest-frequency subband.

6. The signal encoding method of claim 1, wherein the signal components are spectral signal components obtained by converting the input signal from the time domain to the frequency domain.

7. The signal encoding method of claim 1, wherein the input signal is an acoustic signal.

8. The signal encoding method of claim 7, wherein the frequency subbands have been selected to be narrower in the low frequency ranges and broader in the high frequency ranges, to take into account the psychoacoustic characteristics of human hearing.

9. The signal encoding method of claim 1, wherein the encoding method produces a higher ratio of shorter to longer codelengths than if the subband-based signal components are quantized in such a manner as to minimize the total quantization error energy in each subband.

10. A recording medium having encoded signals recorded thereon, the encoded signals operative to control, in part, the operation of a decoding apparatus, the recording medium being prepared by the steps of:

transforming an input signal to a frequency spectrum of signal components;

splitting the signal components of the frequency spectrum into a plurality of frequency subbands;

quantizing the signal components in each subband to generate quantized values;

determining the quantization precision needed for the signal components of each subband, wherein the quantization precision is determined to be higher where the quantized values are larger and lower where the quantized values are smaller;

encoding the quantized values in variable-length codes, wherein the code lengths are set to be longer if the quantization precision value is higher, and shorter if the quantization precision value is lower; and

recording the encoded quantized values onto the recording medium.

11. The recording medium of claim 10, wherein the process of preparing the recording medium further comprises the steps of:

calculating the total number of bits necessary for minimizing the total quantization error energy in each subband;

comparing the total number of bits necessary for minimizing the total quantization error energy in each subband to the total number of available bits; and

setting the range of subband-based signal components which have a short code length to be larger than if the subband-based signal components are quantized in such a manner as to reduce the total quantization error energy in each frequency subband only if the total number of bits necessary for minimizing the total quantization error energy in each subband is larger than the total number of available bits.

12. The recording medium of claim 10, wherein the process of preparing the recording medium further comprises the steps of:

selecting a normalizing coefficient; and

prior to the step of quantizing, normalizing the signal components of each subband.

13. The recording medium of claim 12, wherein the steps of normalizing, quantizing, determining quantization precision and encoding begin in a highest-frequency subband.

14. The recording medium of claim 10, wherein the signal components are spectral signal components obtained by converting the input signal from the time domain to the frequency domain.

15. An encoded data signal recorded on a recording medium, the encoded data signal comprising:

quantized values of signal components in a plurality of frequency subbands of a frequency spectrum;

quantization precision values for the signal components of each subband, wherein the quantization precision is higher where the quantized values are larger and lower where the quantized values are smaller; and

variable-length codes, wherein the code lengths are longer if the quantization precision value is higher, and shorter if the quantization precision value is lower.

16. A signal decoding method, comprising the steps of:

inputting a codestring comprising variable-length codes of quantized spectral signal components, quantization precision information and normalization coefficient information, wherein the quantization precision is higher when the quantized values are larger and lower where the quantized values are smaller, and wherein the codelengths are longer if the quantization precision value is higher and shorter if the quantization precision value is lower;

extracting the codes of spectral signal components from the codestring;

separating the codes of spectral signal components from codes of the quantization precision information and the normalization coefficient information;

decoding the spectral signal components using the quantization precision information and the normalization precision information;

transforming the spectral signal components from the frequency domain to the time domain; and

outputting the result.

17. A signal encoding apparatus for encoding acoustic waveform signals, comprising:

means for inputting time domain signals;

means for transforming an input signal to a frequency spectrum of signal components;

means for splitting the signal components of the frequency spectrum into a plurality of frequency subbands,

means for quantizing the signal components in each subband to generate quantized values;

means for determining the quantization precision needed for the signal components of each subband, wherein the quantization precision is determined to be higher where the quantized values are larger and lower where the quantized values are smaller;

means for encoding the quantized values in variable-length codes, wherein the code lengths are set to be longer if the quantization precision value is higher, and shorter if the quantization precision value is lower; and

means for outputting any of the data generated by the foregoing means.

18. The signal encoding apparatus of claim 17, further comprising:

means for calculating the total number of bits necessary for minimizing the total quantization error energy in each subband;

means for comparing the total number of bits necessary for minimizing the total quantization error energy in each subband to the total number of available bits; and

means for setting the range of subband-based signal components which have a short code length to be larger than if the subband-based signal components are quantized in such a manner as to reduce the total quantization error energy in each frequency subband only if the total number of bits necessary for minimizing the total quantization error energy in each subband is larger than the total number of available bits.

19. The signal encoding apparatus of claim 17, further comprising:

means for selecting a normalizing coefficient; and

means for normalizing the signal components of each subband prior to the operation of the means for quantizing.

20. The signal encoding apparatus of claim 19, wherein the means for normalizing, quantizing, determining quantization precision and encoding begin their operation in a highest-frequency subband.

21. The signal encoding apparatus of claim 19, wherein the means for selecting selects a normalizing coefficient which approximates the maximum absolute value of the spectral signal component of each subband.

22. The signal encoding apparatus of claim 17, wherein the signal components are spectral signal components obtained by converting the input signal from the time domain to the frequency domain.

23. The signal encoding apparatus of claim 17, wherein the input signal is an acoustic signal.

24. The signal encoding apparatus of claim 23, wherein the means for splitting splits the frequency subbands such that the frequency subbands are narrower in the low frequency ranges and broader in the high frequency ranges, to take into account the psychoacoustic characteristics of human hearing.

25. The signal encoding apparatus of claim 17, wherein the means for encoding produces a higher ratio of shorter to longer codelengths than if the subband-based signal components are quantized in such a manner as to minimize the total quantization error energy in each subband.

26. The signal encoding apparatus of any one of claims 17 through 24, further comprising a means for modulating the output signal so that the output signal can be recorded on a recording medium.

27. The signal encoding apparatus of any one of claims 17 through 24, further comprising a means for modulating the output signal so that the output signal can be transmitted over a transmission medium.

28. A signal encoding apparatus for encoding acoustic signals, comprising:

an input terminal for inputting time domain signals;

a conversion circuit for transforming an input signal to a frequency spectrum of signal components and splitting the signal components into a plurality of frequency subbands;

a signal component encoding circuit for quantizing the signal components in each subband to generate quantized values and for determining the quantization precision needed for the signal components of each subband, wherein the quantization precision is determined to be higher where the quantized values are larger and lower where the quantized values are smaller;

a codestring generating circuit for encoding the quantized values in variable-length codes, wherein the code lengths are set to be longer if the quantization precision value is higher, and shorter if the quantization precision value is lower; and

an output terminal for outputting any of the data generated by the conversion circuit, the signal component encoding circuit or the codestring generating circuit.

29. The signal encoding apparatus of claim 28, wherein the conversion circuit comprises:

a band-splitting filter; and

two forward orthogonal transform circuits.

30. The signal encoding apparatus of claim 28, wherein the encoding circuit comprises:

a normalizing circuit, for normalizing the spectral signal components from one subband to another; and

a quantization precision decision circuit, which quantizes the normalized values based on the quantization precision, as computed from the spectral signal components.

31. The signal encoding apparatus of claim 28, wherein the signal component encoding circuit comprises:

a quantization error energy calculating circuit for calculating the total number of bits necessary for minimizing the total quantization error energy in each subband;

a comparing circuit for comparing the total number of bits necessary for minimizing the total quantization error energy in each subband to the total number of available bits; and

a range setting circuit for setting the range of subband-based signal components which have a short code length to be larger than if the subband-based signal components are quantized in such a manner as to reduce the total quantization error energy in each frequency subband only if the total number of bits necessary for minimizing the total quantization error energy in each subband is larger than the total number of available bits.

32. The signal encoding apparatus of claim 28, wherein the signal component encoding circuit comprises:

a normalization coefficient selection circuit for selecting a normalizing coefficient; and

a normalizing circuit for normalizing the signal components of each subband prior to the operation of the means for quantizing.

33. The signal encoding apparatus of claim 32, wherein the signal component encoding circuit begins its operation in a highest-frequency subband.

34. The signal encoding apparatus of claim 32, wherein the signal component encoding circuit selects a normalizing coefficient which approximates the maximum absolute value of the spectral signal component of each subband.

35. The signal encoding apparatus of claim 28, wherein the signal components are spectral signal components obtained by converting the input signal from the time domain to the frequency domain.

36. The signal encoding apparatus of claim 28, wherein the input signal is an acoustic signal.

37. The signal encoding apparatus of claim 36, wherein the conversion circuit splits the frequency subbands such that the frequency subbands are narrower in the low frequency ranges and broader in the high frequency ranges, to take into account the psychoacoustic characteristics of human hearing.

38. The signal encoding apparatus of claim 28, wherein the signal component encoding circuit produces a higher ratio of shorter to longer codelengths than if the subband-based signal components are quantized in such a manner as to minimize the total quantization error energy in each subband.

39. The signal encoding apparatus of any one of claims 28 through 37, further comprising a modulating circuit for modulating the output signal so that the output signal can be recorded on a recording medium.

40. The signal encoding apparatus of any one of claims 28 through 37, further comprising a modulating circuit for modulating the output signal so that the output signal can be transmitted over a transmission medium.

41. A signal decoding apparatus, comprising:

means for inputting a codestring comprising variable-length codes of quantized spectral signal components, quantization precision information and normalization coefficient information, wherein the quantization precision is higher when the quantized values are larger and lower where the quantized values are smaller, and wherein the codelengths are longer if the quantization precision value is higher and shorter if the quantization precision value is lower;

means for extracting codes of spectral signal components from the codestring and separating the spectral signal components from codes of the quantization precision information and the normalization coefficient information;

means for decoding the spectral signal components using the quantization precision information and the normalization precision information;

means for transforming the spectral signal components from the frequency domain to the time domain; and

means for outputting the decoded signal.

42. A signal decoding apparatus, comprising:

an input terminal for inputting a codestring comprising variable-length codes of quantized spectral signal components, quantization precision information and normalization coefficient information, wherein the quantization precision is higher when the quantized values are larger and lower where the quantized values are smaller, and wherein the codelengths are longer if the quantization precision value is higher and shorter if the quantization precision value is lower;

a codestring resolution circuit, in which codes of spectral signal components are extracted from the codestring and separated from codes of the quantization precision information and the normalization coefficient information;

a signal component decoding circuit which decodes the spectral signal components using the quantization precision information and the normalization precision information;

an inverse transform circuit which transforms the spectral signal components from the frequency domain to the time domain; and

an output terminal.