Lossless coding method for waveform data

• US 6,664,913 B1
• Filed: 05/17/1999
• Issued: 12/16/2003
• Est. Priority Date: 05/15/1995
• Status: Expired due to Term

• Current Assignee: Dolby Laboratories Licensing Corporation (Dolby Laboratories Incorporated)
• Original Assignee: Dolby Laboratories Licensing Corporation (Dolby Laboratories Incorporated)
• Inventors: Craven, Peter G., Gerzon, Michael A.
• Primary Examiner(s): Tokar, Michael
• Assistant Examiner(s): Mai, Lam T.
• Application Number:
  US08/952,743
• Time in Patent Office: 1,674 Days
• Field of Search: 341/200, 341/143, 341/76, 341/77, 341/138, 704/211, 704/212, 704/222, 704/230, 375/242, 375/243, 375/245
• US Class Current: 341/200
• CPC Class Codes: H03M 7/3046 adaptive, e.g. adaptive differential pulse code modulation [ADPCM]

Abstract

In a method of lossless processing of an integer value signal in a prediction filter which includes a quantiser, a numerator of the prediction filter is implemented prior to the quantiser and a denominator of the prediction filter is implemented recursively around the quantiser to reduce the peak data rate of an output signal. In the lossless processor, at each sample instant, an input to the quantiser is jointly responsive to a first sample value of a signal input to the prediction filter, a second sample value of a signal input to the prediction filter at a previous sample instant, and an output value of the quantiser at a previous sample incident. In a preferred embodiment, the prediction filter includes noise shaping for affecting the output of the quantiser.

Claims

1. An invertible transformation signal processor for converting input integer-valued PCM n-channel waveform data to output integer-valued PCM n-channel waveform data, wherein the processor comprises:

an n-channel input for said input integer-valued PCM n-channel waveform data;

an n-channel output for said output integer-valued PCM n-channel waveform data;

an n-channel summing node;

an n-channel differencing node;

an n-channel quantiser rounding operation rounding to near or nearest integer values;

a first n×

n matrix filter means, each term of which incorporates delay;

a second n×

n matrix filter means, each term of which incorporates delay; and

,

a first and a second n×

n matrix quantiser,

wherein the n-channel input is coupled to the summing node via said first n×

n matrix quantiser, the n-channel input is coupled to the differencing node via the first n×

n matrix filter, the n-channel output is coupled to the differencing node via the second n×

n matrix filter, the n-channel differencing node is coupled to the n-channel quantiser, the n-channel quantiser is coupled to the n-channel summing node, and the n-channel summing node is coupled to the n-channel output via the second n×

n matrix quantiser.

2. A signal processor according to claim 1, in which the first and second matrix filters are matrix FIR filters.

3. A signal processor according to claim 2, in which said first and second n×

n matrix FIR filters have coefficients which have finite fractional precision.

4. A lossless matrix quantizer for quantizing waveform data comprising signals for two or more channels, wherein the lossless matrix quantizer comprises:

a first input and one or more second inputs, wherein the first input and the one or more second inputs each receives an integer-valued signal for a respective channel;

a single-channel quantizer having a quantizer input coupled to the one or more second inputs and having a quantizer output;

a first output coupled to the quantizer output that provides an output signal representing a linear combination of the signal received at the first input and a quantization of each signal received at a respective second input, wherein the first output provides a transformation of the integer-valued signal received at the first input that is invertible for each sample of the output signal given knowledge of previous samples received at the first input and current and previous samples received at the one or more second inputs; and

one or more second outputs each coupled to a respective second input.

5. A lossless matrix quantizer according to claim 4 that comprises a summing node, wherein the summing node has an output coupled to the quantizer input, and has inputs coupled to the first input and the one or more second inputs such that the quantizer input is responsive to a sum or difference of the signal received at the first input and a linear combination of the signals received at the one or more second inputs.

6. A lossless matrix quantizer according to claim 5 that comprises a summing node, wherein the summing node has inputs coupled to the quantizer output and the first input, and has an output coupled to the first output.

7. A lossless matrix quantizer according to claim 5, wherein the quantizer input receives an autodither signal comprising a sequence of dither samples, a respective dither sample derived from one or. more previous samples of a signal at the quantizer output and/or one or more previous samples of a signal at the quantizer input.

8. An apparatus comprising a cascade of plural lossless matrix quantizers, wherein each respective lossless matrix quantizer is according to claim 4.

9. An apparatus according to claim 8 for quantizing waveform data comprising a signal for each of N channels, wherein the cascade of lossless matrix quantizers provides a set of N signal paths and the lossless matrix quantizers in the cascade are coupled to one another such that each of the N signal paths includes the first input of at least one lossless matrix quantizer.

10. An apparatus according to claim 8 for quantizing waveform data comprising a signal for each of N channels, wherein the cascade of lossless matrix quantizers provides a set of N signal paths and the lossless matrix quantizers in the cascade are coupled to one another such that each of the N signal paths includes the first input of no more than one lossless matrix quantizer.

11. A method for using a lossless matrix quantizer to quantize waveform data comprising signals for two or more channels, wherein the method comprises:

receiving at each of a first input and one or more second inputs an integer-valued signal for a respective channel;

applying a single-channel quantizer having a quantizer output to one or more signal coupled from the one or more second inputs;

providing an output signal at a first output coupled to the quantizer output that represents a linear combination of the signal received at the first input and a quantization of each signal received at a respective second input, wherein the first output provides a transformation of the integer-valued signal received at the first input that is invertible for each sample of the output signal given knowledge of previous samples received at the first input and current and previous samples received at the one or more second inputs; and

coupling a signal from each second input to a respective second output.

12. A method according to claim 11 that comprises coupling the signals received at the first input and the one or more second inputs to a summing node and coupling the output of the summing node to the quantizer such that the quantizer is responsive to a sum or difference of the signal received at the first input and a linear combination of the one or more signals received at the one or more second inputs.

13. A lossless matrix quantizer according to claim 11 that comprises summing the quantizer output and the signal received at the first input and coupling the result to the first output.

14. A method according to claim 11, wherein the quantizer is also applied to an autodither signal comprising a sequence of dither samples, wherein a respective dither sample is derived from one or more previous samples of a signal at the quantizer output and/or one or more previous samples of a signal to which the quantizer is applied.

15. A method that comprises performing two or more times the steps of claim 11 for using a cascade of a plurality of lossless matrix quantizers.

16. A method according to claim 15 for quantizing waveform data comprising a signal for each of N channels, wherein the cascade of lossless matrix quantizers provides a set of N signal paths and the lossless matrix quantizers in the cascade are coupled to one another such that each of the N signal paths includes the first input of at least one lossless matrix quantizer.

17. A method according to claim 15 for quantizing waveform data comprising a signal for each of N channels, wherein the cascade of lossless matrix quantizers provides a set of N signal paths and the lossless matrix quantizers in the cascade are coupled to one another such that each of the N signal paths includes the first input of no more than one lossless matrix quantizer.