Variable frame rate, fixed bit rate vocoding method

• US 4,852,179 A
• Filed: 10/05/1987
• Issued: 07/25/1989
• Est. Priority Date: 10/05/1987
• Status: Expired due to Fees

• Current Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Fette, Bruce A.
• Primary Examiner(s): Salce, Patrick R.
• Assistant Examiner(s): VOELTZ, EMANUEL T
• Application Number:
  US07/104,698
• Time in Patent Office: 659 Days
• Field of Search: 381/29-53, 364/513.5
• US Class Current: 704/230
• CPC Class Codes: G10L 19/04 using predictive techniques
• CPC Class Codes: H03M 7/3082 Vector coding for television signals, see H04N19/94

Abstract

A method of operating a vocoder so that a variable frame rate results while maintaining a constant bit rate is disclosed. A base frame rate remains constant. However, spectral change is measured to determine an appropriate subframe rate within the base frame. Only one of a plurality of separate vector quantization processes is selected in response to spectral change for performance of a vector quantization operation on LPC spectrum coefficients. Each of the plurality of vector quantization operations utilizes its own codebook that contains a different quantity of reference patterns from that contained in the other codebooks. Each vector quantization operation produces a reference pattern descriptor code that contains a different number of bits from that produced by the other vector quantization processes. Vector quantization operations producing smaller, less spectrally accurate outputs are selected when more subframes are included with a base frame, and vector quantization operations producing larger, more spectrally accurate outputs are selected when fewer subframes are included within a base frame.

Claims

1. A method of operating a vocoder to compress a multiplicity of voice data samples for application to a constant rate channel, said method comprising the steps of:

selecting one of a plurality of subframe rates;

analyzing at lest a portion of the samples to produce a predictive code representative of the analyzed samples;

quantizing the predictive code into a quantized code having a bit length defined by the selected subframe rate and a rate parameter of the constant rate channel; and

said steps of selecting, analyzing and quantizing being performed at a bit rate of said constant rate channel.

2. A method as claimed in claim 1 additionally comprising the step a base frame of data to be applied to the constant rate channel at a fixed base frame rate and to include information which remains constant throughout the base frame in addition to a quantized code for at least one subframe.

3. A method as claimed in claim 2 additionally comprising the step of repeating said analyzing and quantizing steps when the selected subframe rate permits more than one subframe to occur during a base frame.

4. A method as claimed in claim 1 wherein said selecting step comprises the steps of:

measuring spectral change occurring between the multiplicity of voice data samples and a prior multiplicity of voice data samples; and

modulating subframe rate in response to the spectral change.

5. A method as claimed in claim 4 wherein said modulating step comprises the step of increasing subframe rate in response to increasing spectral change.

6. A method as claimed in claim 1 wherein said quantizing step comprises the step of performing a vector quantization of the predictive code.

7. A method as claimed in claim 6 wherein said quantizing step comprises the steps of:

making first and second vector quantization codebooks available for searching;

choosing only one of the first and second vector quantization codebooks in response to said selecting step;

searching within the chosen one from said choosing step of the first and second vector quantization codebooks for a nearest matching reference pattern to the predective code; and

retrieving a reference pattern descriptor code which uniquely describes the nearest matching reference pattern.

8. A method as claimed in claim 7 wherein said making step comprises the step of populating the first vector quantization codebook with a quantity of reference patterns that differs from the quantity of reference patterns that populate the second vector quantization codebook so that a first reference pattern descriptor code retrievable from the first codebook for the predective code is described utilizing a different number of bits than describes a second reference pattern descriptor code retrievable from the second codebook for the predective code.

9. A method as claimed in claim 7 wherein said analyzing step performs a linear predictive coding of the multiplicity of voice data samples to generate excitation, energy, and spectrum data, and said searching step comprises the step of transforming the spectrum data into an N-bit code wherein the chosen codebook contains no more than 2^N unique reference patterns.

10. A method of operating a vocoder to decompress a base frame of data which represents a multiplicity of voice data samples and which is received from a constant rate channel, said method comprising the steps of:

obtaining information describing one of a plurality of subframe rates from the constant rate base frame of data;

separating the base frame of data into a least one subframe of data in response to the subframe rate from said obtaining step;

expanding each subframe of data from said separating step into an independent predictive code;

synthesizing the multiplicity of voice data samples from the predictive code for each subframe; and

said steps of obtaining, separating, expanding and synthesizing being performed at a bit rate of said constant rate channel.

11. A method as claimed in claim 10 wherein said expanding step comprises the step of performing an inverse vector quantization of each subframe of data.

12. A method as claimed in claim 11 wherein said expanding step comprises the steps of:

making first and second vector quantization codebooks available for the inverse vector quantization operations;

choosing only one of the first and second vector quantization codebooks in response to the subframe rate from said obtaining step;

operating with the chosen one from said choosing step of the first and second vector quantization codebooks to generate reference patterns that uniquely match the predictive code for each subframe of data.

13. A method as claimed in claim 12 wherein said making step comprises the step of populating the first vector quantization codebook with a quantity of reference patterns that differs from the quantity of reference patterns that populate the second vector quantization codebook so that a first reference pattern descriptor code utilized with the first vector quantization codebook is described utilizing a different number of bits than describes a second reference pattern descriptor code utilized with the second codebook.

14. A method as claimed in claim 12 wherein the chosen codebook from said choosing step contains no more than 2^N reference patterns, wherein the predictive code describes excitation, energy, and spectrum data, wherein said synthesizing step performs a linear predictive coding of the predictive code to generate the multiplicity of voice data samples, and said operating step comprises the step of transforming only an N-bit portion of the subframe of data into the spectrum data portion of the predictive code.

15. A method of operating a vocoder to compress a multiplicity of voice data samples for application to a fixed rate channel, said method comprising the steps of:

measuring spectral change occurring between the multiplicity of voice data samples and a prior multiplicity of voice data samples;

modulating a subframe rate in response to the spectral change so that increasing spectral change causes increasing subframe rate;

performing a linear predictive coding of at least a portion of the multiplicity of voice data samples to generate excitation, energy and spectrum data;

making first and second vector quantization codebooks available for searching;

choosing one of the first and second vector quantization codebooks in response to the subframe rate of said modulating step;

searching within only the one codebook chosen in said choosing step for a nearest matching reference pattern to the spectrum data of said performing step; and

retrieving a reference pattern descriptor code which uniquely describes the nearest matching reference pattern, wherein the reference pattern descriptor code contains N bits and the one codebook chosen in said choosing step contains less than 2^N unique reference patterns.

16. A method as claimed in claim 15 additionally comprising the step of repeating said performing, searching, and retrieving steps when the subframe rate of said modulating step permits more than one subframe to occur during a base frame.