Vector-quantizing device having a capability of adaptive updating of code book

• US 5,524,170 A
• Filed: 06/15/1994
• Issued: 06/04/1996
• Est. Priority Date: 08/30/1991
• Status: Expired due to Fees

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Okazaki, Koji, Matsuo, Naoshi
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Doerrler, Michelle
• Application Number:
  US08/261,208
• Time in Patent Office: 720 Days
• Field of Search: 395/2.28-2.39, 381/29-40, 358/133, 358/135
• US Class Current: 704/222
• CPC Class Codes: G06T 9/008 Vector quantisation
• CPC Class Codes: G10L 19/038 Vector quantisation, e.g. TwinVQ audio
• CPC Class Codes: H03M 7/3082 Vector coding for television signals, see H04N19/94

Abstract

A vector-quantizing device comprises a first code book for storing a plurality of code vectors at respective addresses; a first addressing unit for selecting a code vector stored in the first code; a synthesis unit supplied with the code vector from the first code book for producing a synthetic signal that simulates an input signal; an evaluation unit supplied with the input signal and further with the synthetic signal from the synthesis means for evaluating a difference therebetween and searching for a code vector that minimizes the difference. There, the evaluation unit produces the first index data that specifies a code vector wherein the difference is minimized as a coded output. In addition, the vector-quantizing device includes a second code book for storing a plurality of code vectors at respective addresses; a second addressing unit for selecting a code vector stored in the second code book: and an update unit for counting the frequency of selection of the code vectors for each of the code vectors stored in the first code book. There, the update unit is for replacing a code vector in the first code book, when the frequency of use of the code vector is below a predetermined threshold, with a code vector in the second code book.

Claims

1. A vector-quantizing device for coding an input signal by a vector-quantization process, comprising:

a first code book for storing a plurality of code vectors at respective addresses;

first addressing means supplied with first index data for selecting a code vector stored only in said first code book in response to said first index data;

synthesis means supplied with only said selected code vector from said first code book for producing a synthetic signal that simulates said input signal;

evaluation means supplied with said input signal and further with said synthetic signal from said synthesis means for evaluating a difference therebetween, said evaluation means searching for a code vector that minimizes said difference in said first code book, transmitting said first index data that specifies a code vector wherein said difference is minimized as a coded output and producing a count signal each time a code vector in said first code book is selected;

a second code book for storing a plurality of code vectors at respective addresses;

second addressing means supplied with second index data for selecting a code vector stored in said second code book;

counting means for counting a frequency of selection of the code vectors for each of said code vectors only stored in said first code book using said count signal; and

update means for producing said second index data and replacing a code vector stored in said first code book and addressed by said first index data, when said frequency of selection of said code vector stored in said first code book and addressed by said first index data is below a predetermined threshold, with a code vector stored in said second code book and addressed by said second index data.

2. A vector-quantizing device as claimed in claim 1 in which said update means comprises:

threshold means supplied with the result of counting from said counting means for each of said code vectors in said first code book and further with a predetermined threshold for identifying the code vectors in said first code book that are selected with a frequency smaller than said predetermined threshold;

second index data generation means for producing said second index data;

and data transfer means for transferring the code vector stored in said second code book, when addressed by said second index data, to said first code book such that the code vector in said first code book and identified by said threshold means as being the code vector of which frequency of selection is smaller than said predetermined threshold, is overwritten by said code vector transferred from said second code book.

3. A vector-quantizing device as claimed in claim 2, wherein said counting means comprises a plurality of counters provided in correspondence to each index of the code vectors in said first code book, and said evaluation means supplies said first index data to said counting means as said count signal.

4. A vector-quantizing device as claimed in claim 2, wherein said data transfer means comprises a buffer amplifier for reading out the code vector from said second code book via said second addressing means and for writing the same into said first code book via said first addressing means.

5. A vector-quantizing device as claimed in claim 2, wherein said second index data generation means comprises an index counter for counting a number of times the code vectors in said first code book is overwritten and produces said number as said second index data.

6. A vector-quantizing device as claimed in claim 2, wherein said second index data generation means comprises an index counter running at a predetermined rate and producing a counted number as said second index data.

7. A vector-quantizing device as claimed in claim 2, wherein said second index data generation means comprises a random number generator that produces a random number as said second index data.

8. A vector-quantizing device as claimed in claim 1, wherein said device further comprises a training circuit supplied with said input for signal for modifying said code vectors stored in said first code book such that said difference evaluated by said evaluation means is minimized.

9. A coder that uses a vector-quantizing device comprising:

an encoder; and

a decoder;

said encoder comprising;

a first code book for storing a plurality of code vectors at respective addresses;

first addressing means supplied with first index data for selecting a code vector stored only in said first code book in response to said first index data;

synthesis means supplied only with said selected code vector from said first code book for producing a synthetic signal that simulates said input signal;

evaluation means supplied with said input signal and further with said synthetic signal from said synthesis means for evaluating a difference therebetween, said evaluation means searching for a code vector that minimizes said difference in said first code book, transmitting said first index data that specifies a code vector wherein said difference is minimized as a coded output and producing a count signal each time a code vector in said first code book is selected;

a second code book for storing a plurality of code vectors at respective addresses;

second addressing means supplied with second index data for selecting a code vector stored in said second code book;

counting means for counting the frequency of selection of the code vectors for each of said code vectors only stored in said first code book using said count signal; and

update means for producing said second index data and replacing a code vector stored in said first code book and addressed by said first index data, when the frequency of selection of said code vector stored in said first code book and addressed by said first index data is below a predetermined threshold, with a code vector stored in said second code book and addressed by said second index data; and

said decoder comprising;

a third code book for storing a plurality of code vectors at respective addresses;

third addressing means supplied with said coded output from said evaluation means as third index data for selecting a code vector stored in said third code book in response to said third index data;

second synthesis means supplied with said selected code vector from said third code book for producing a synthetic signal that simulates said input signal as a reproduced signal;

a fourth code book for storing a plurality of code vectors at respective addresses;

fourth addressing means supplied with fourth index data for selecting a code vector stored in said fourth code book; and

second update means for counting the frequency of selection of the code vectors for each of said code vectors stored in said third code book, said update means producing said third index data and said fourth index data for replacing a code vector stored in said third code book and addressed by said third index data, when the frequency of use of said code vector is below a predetermined threshold, with a code vector stored in said fourth code book and addressed by said fourth index data.

10. A coder as claimed in claim 9, wherein said encoder further comprises:

a fifth code book for storing gain code values representing the gain at respective addresses;

fifth addressing means supplied with fifth index data from said evaluation means for selecting a gain code value in said fifth code book; and

gain amplifier means supplied with a shape code vector from said first code book via said first addressing means and a gain code value from said fifth code book via said fifth addressing means for amplifying said shape code vector supplied from said first code book with a gain specified by said gain code value from said fifth code book, said coder outputting said fifth index data as part of said coded output.

11. A coder as claimed in claim 10, wherein said decoder further comprises:

a sixth code book for storing gain code values representing the gain at respective addresses; and

sixth addressing means supplied with sixth index data from said encoder as said coded output for selecting a gain code value stored in said sixth code book, and gain amplifying means for amplifying said code vector read out from said third code book by said third addressing means with a gain specified by said gain code value read out from said sixth code book by said sixth addressing means.

12. A method for encoding an input signal with a vector-quantizing process, comprising the steps of:

quantizing the input signal usinq only a code vector in a first code book;

detecting a count frequency of use of the code vector stored in the first code book;

identifying code vectors of the first code book of which a frequency of use is smaller than a predetermined threshold; and

replacing the code vectors of the first code book thus identified with another code vector stored in a second code book.