Resisting the effects of channel noise in digital transmission of information

• US 4,791,654 A
• Filed: 06/05/1987
• Issued: 12/13/1988
• Est. Priority Date: 06/05/1987
• Status: Expired due to Term

• Current Assignee: American Telephone & Telegraph Company (AT&T, Inc.), Bell Telephone Laboratories, Inc. (Nokia Corporation)
• Original Assignee: American Telephone & Telegraph Company (AT&T, Inc.)
• Inventors: De Marca, Jose R. B., Jayant, Nuggehally S.
• Primary Examiner(s): Griffin, Robert L.
• Assistant Examiner(s): Chin, Stephen
• Application Number:
  US07/058,728
• Time in Patent Office: 557 Days
• Field of Search: 375/27.30, 375/122, 358/13, 358/133, 358/134, 340/347 AD, 381/29, 381/30, 381/34, 381/35
• US Class Current: 375/241
• CPC Class Codes: G10L 19/038 Vector quantisation, e.g. TwinVQ audio
• CPC Class Codes: G10L 19/06 Determination or coding of the spectral characteristics, e.g. of the short-term prediction coefficients
• CPC Class Codes: H03M 7/3082 Vector coding for television signals, see H04N19/94
• CPC Class Codes: H04N 19/94 Vector quantisation

Abstract

A method for resisting the effects of channel noise in the digital transmission of information by means of the technique known as vector quantization, in which the codebook for binary index code assignment is generated by picking a vector quantized codeword with high probability and low perceptually-related distance from a required group of nearest neighbors, assigning that codeword and those neighbors binary index codes differing only in one bit, repeating the steps just outlined for assigned binary index codes to residual codewords until the lasts assignments must be made arbitrarily. Applications include transmission of speech by coded LPC parameters and transmission of image intensity or chrominance blocks for visual images. Such systems also make possible and are attractive for memory-efficient storage of such signals.

Claims

1. A method of the type employing the assigning of binary index codes to codewords representing vector quantized information, for transmission in a communication system

characterized in that the assigning includes the steps of

picking a first codeword having a high probability of occurrence and low inter-codeword distance during system operation,

assigning a first binary index code to the first codeword,

finding the nearest neighbor codewords, in terms of some logical proximity of the nearest neighbor codewords to the first codeword,

assigning to at least some of the nearest neighbor codewords binary index codes differing from the first binary index code only in any one bit,

repeating the foregoing steps as many times as needed for successive residual high probability codewords, while avoiding duplicate assignments until no suitable group of neighboring codewords remain, and

assigning residual binary index codes to residual codewords.

2. A method according to claim 1 characterized in that

the finding step includes measuring the distance or proximity of the nearest neighbor codewords in terms of some perceptually-related similarity of the underlying codewords, such that the decoded information will be perceived to be resistant to noise occurring during transmission in the communication system, provided that the stored assignment information is employed in such a system in which noise typically affects only one bit in any affected binary index code.

3. A method according to claim 2 characterized in that the communication system is a linear predictive coding (LPC) type of speech transmission system, and the finding step includes measuring the underlying sound similarities of the groups of LPC parameter samples, where each group of samples has been vector quantized together.

4. A method according to claim 2 characterized in that the communication system is an image transmission type of system, and the finding step includes measuring the underlying spot pattern similarities, as perceived, of groups of spots of image intensity or chrominance, where each group of spots has been vector quantized together.

5. A method according to claim 3 or 4 characterized in that the picking step includes measuring the probability of occurrence ofa codeword as well as the total distance of B-nearest-neighbor codewords, where distances are measured in terms of perceptually-related proximity.

6. Communication equipment of the type including means for assigning binary index codes to codewords representing vector quantized information to be communicated,

characterized in that the means for assigning includes

means for picking a first codeword having a high probability of occurrence and low inter-codeword distance,

means for assigning a first binary index code to the first codeword,

means for finding the nearest neighbor codewords, in terms of some logical proximity of the nearest neighbor codewords to the first codeword,

means for assigning to at least some of the nearest neighbor codewords binary index codes differing from the first binary index code only in any one bit,

means for processing successive residual high probability codewords through the preceding three means, including means for avoiding duplicate assignments, and

means for assigning residual binary index codes to residual codewords having no suitable residual nearest neighbor codewords.