Distance and distortion estimation method and apparatus in channel optimized vector quantization

• US 6,321,193 B1
• Filed: 01/27/1999
• Issued: 11/20/2001
• Est. Priority Date: 01/27/1998
• Status: Expired due to Term

• Current Assignee: Telefonaktiebolaget LM Ericsson
• Original Assignee: Telefonaktiebolaget LM Ericsson
• Inventors: Nyström, Johan, Hagen, Roar, Svensson, Tomas, Minde, Tor Björn
• Primary Examiner(s): Korzuch, William
• Assistant Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US09/238,386
• Time in Patent Office: 1,028 Days
• Field of Search: 704/221, 704/222, 704/223, 704/226, 704/230
• US Class Current: 704/222
• CPC Class Codes: H03M 7/3082 Vector coding for television signals, see H04N19/94

Abstract

A channel optimized vector quantization apparatus includes a device for weighting a sample vector x by a weighting matrix A and a device for weighting a set of code book vectors ĉ_r by a weighting matrix B. Device form a set of distance measures {d_w(Ax,Bĉ_r)} representing the distance between the weighted sample vector Ax and each weighted code book vector Bĉ_r. Other device form a set of distortion measures {_αi(x)} by multiplying each distance measure by a channel transition probability P_r|i that an index r has been received at a decoder when an index i has been sent from an encoder and adding together these multiplied distance measures for each possible index r. Finally device determine an index i_min corresponding to the smallest distortion measure _αi(x) and represents the sample vector by this index i_min.

Claims

1. A distance estimation method in channel optimized vector quantization, comprising the steps of:

forming a sample vector x by sampling an input signal;

weighting the sample vector x by a first weighting matrix A;

weighting a code book vector ĉ

_r by a second weighting matrix B different from said first weighting matrix A; and

forming a distance measure d_w (Ax,B<sup>ĉ

_r</sup>) representing the distance between said weighted sample vector Ax and said weighted code book vector B<sup>ĉ

_r</sup>.

2. The method of claim 1, wherein said sample vector x and said code book vector ĉ

_r are of the same length.

3. The method of claim 1, wherein said distance measure d_w(Ax,Bĉ

_r) is a weighted squared Euclidean distance measure.

4. The method of claim 1, wherein only one of said first weighting matrix A and said second weighting matrix B is different from the identity matrix.

5. The method of claim 1, wherein each weighting matrix A,B is sample vector independent.

6. The method of claim 1, wherein each weighting matrix A,B is time independent.

7. The method of claim 1, wherein each weighting matrix A,B is constant.

8. A distortion estimation method in channel optimized vector quantization, comprising the steps of:

forming a sample vector x by sampling an input signal;

weighting the sample vector x by a first weighting matrix A;

weighting a set of code book vectors ĉ

_r by a second weighting matrix B different from said first weighting matrix A;

forming a set of distance measures d_w(Ax,B<sup>ĉ

_r</sup>) representing the distance between said weighted sample vector Ax and each weighted code book vector B<sup>ĉ

_r</sup>;

forming a distortion measure α

_i(x) by multiplying each distance measure by a predetermined channel transition probability p_r|i that an index r has been received at a decoder when an index i has been sent from an encoder and adding together said multiplied distance measures for each possible index r.

9. A channel optimized vector quantization method, comprising the steps of:

forming a sample vector x by sampling an input signal;

weighting the sample vector x by a first weighting matrix A;

weighting a set of code book vectors ĉ

_r by a second weighting matrix B different from said first weighting matrix A;

forming a set of distance measures d_w(Ax,B<sup>ĉ

_r</sup>) representing the distance between said weighted sample vector Ax and each weighted code book vector B<sup>ĉ

_r</sup>;

forming a set of distortion measures {α

_i(x)} by multiplying each distance measure by a predetermined channel transition probability p_r|i that an index r has been received at a decoder when an index i has been sent from an encoder and adding together said multiplied distance measures for each possible index r;

determining an index i_min corresponding to the smallest distortion measure α

_i(x); and

representing said sample vector by this index i_min.

10. A channel optimized vector quantization method, comprising the steps of:

forming a sample vector x by sampling an input signal;

calculating a set of distortion measures {α

_i′

(x)} in accordance with the expression

11. A distortion estimation method in channel optimized vector quantization, comprising the steps of:

forming a sample vector x by sampling an input signal;

calculating a distortion measure α

_i′

(x) in accordance with the expression

12. The method of claim 11, wherein said expected i:

th reconstruction vector is defined as ${\mathrm{\mu <br><br>}}_i=\underset{r=0}{\overset{M-1}{\sum <br><br>}}<br><br>{\hat{c}}_r·<br><br>p_{r|i}$

where

ĉ

_r is the r;

th code book vector, and

p_r|i is a predetermined channel transition probability that an index r has been received at a decoder when an index i has been sent from an encoder.

13. The method of claim 11 wherein said conditional i:

th code book correlation matrix is defined as ${\mathrm{\Phi <br><br>}}_i=\underset{r=0}{\overset{M-1}{\sum <br><br>}}<br><br>{\hat{c}}_r<br><br>{\hat{c}}_r^T·<br><br>p_{r|i}$

where

ĉ

_r is the r;

th code book vector, and p1 p_r|i is a predetermined channel transition probability that an index r has been received at a decoder when an index i has been sent from an encoder.

14. The method of claim 11, further comprising:

pre-computing and storing the quantities −

2<sub>μ

i</sub> and Φ

_i; and

retrieving these stored quantities each time a distortion measure α

_i′

(x) is to be calculated for a given sample vector x.

15. The method of claim 14, wherein only said code book vector weighting matrix B is different from the identity matrix.

16. The method of claim 14, wherein said sample vector weighting matrix A and said code book weighting matrix B are the same matrix.

17. The method of claim 14, wherein said sample vector weighting matrix A is diagonal.

18. The method of claim 14, wherein said code book vector weighting matrix B is diagonal.

19. The method of claim 11, wherein only said sample vector weighting matrix A is different from the identity matrix, said method further comprising:

pre-computing and storing the quantities −

2<sub>μ

i</sub> and 1^TΦ

_i1; and

retrieving these stored quantities each time a distortion measure α

_i′

(x) is to be calculated for a given sample vector x.

20. The method of claim 11, wherein said code book vector weighting matrix B constant and different from the identity matrix, said method further comprising:

pre-computing and storing the quantities −

2B<sub>μ

i</sub> and 1^T(Φ

_i{circle around (×

)}(B^TB))1; and

retrieving these stored quantities each time a distortion measure α

_i′

(x) is to be calculated for a given sample vector x.

21. The method of claim 11, wherein said sample vector weighting matrix A and said code book vector weighting matrix B are both constant and different from the identity matrix, said method further comprising

pre-computing and storing the quantities −

2A^TB<sub>μ

i</sub> and 1^T(Φ

_i{circle around (×

)}(B^TB))1; and

retrieving these stored quantities each time a distortion measure α

_i′

(x) is to be calculated for a given sample vector x.

22. A distance estimation apparatus in channel optimized vector quantization, comprising:

means for forming a sample vector x by sampling an input signal;

means for weighting the sample vector x by a first weighting matrix A;

means for weighting a code book vector ĉ

_i by a second weighting matrix B different from said first weighting matrix A; and

means for forming a distance measure d_w (Ax,Bĉ

^_i) representing the distance between said weighted sample vector Ax and said weighted code book vector B<sup>ĉ

_i</sup>.

23. A distortion estimation apparatus in channel optimized vector quantization, comprising:

means for forming a sample vector x by sampling an input signal;

means for weighting the sample vector x by a first weighting matrix A;

means for weighting a set of code book vectors ĉ

_r by a second weighting matrix B different from said first weighting matrix A;

means for forming a set of distance measures d_w (Ax,B<sup>ĉ

_r</sup>) representing the distance between said weighted sample vector Ax and each weighted code book vector B<sup>ĉ

_r</sup>;

means for forming a distortion measure by multiplying each distance measure by a predetermined channel transition probability p_r|i that an index r has been received at a decoder when an index i has been sent from an encoder and adding together said multiplied distance measures for each possible index r.

24. A channel optimized vector quantization apparatus, comprising:

means for forming a sample vector x by sampling an input signal;

means for weighting the sample vector x by a first weighting matrix A;

means for weighting a set of code book vectors ĉ

_r by a second weighting matrix B different from said first weighting matrix A;

means for forming a set of distance measures {d_w (Ax,B<sup>ĉ

_r</sup>)} representing the distance between said weighted sample vector Ax and each weighted code book vector B<sup>ĉ

_r</sup>;

means for forming a set of distortion measures {α

_i(x)} by multiplying each distance measure by a predetermined channel transition probability P_r|i that an index r has been received at a decoder when an index i has been sent from an encoder and adding together said multiplied distance measures for each possible index r;

means for determining an index i_min corresponding to the smallest distortion measure α

_i(x) and representing said sample vector by this index i_min.

25. A distortion estimation apparatus in channel optimized vector quantization, comprising:

means for forming a sample vector x by sampling an input signal; and

means for calculating a distortion measure α

_i′

(x) in accordance with the expression

26. A channel optimized vector quantization apparatus, comprising:

means for forming a sample vector x by sampling an input signal;

means for calculating a set of distortion measures {α

_i′

(x)} in accordance with the expression

27. The apparatus of claim 26, comprising:

means for storing the pre-computed quantities −

2<sub>μ

i</sub> and Φ

_i and for retrieving these stored quantities each time a distortion measure α

_i′

(x) is to be calculated for a given sample vector x.