Browse Prior Art Database

Gain Control to Improve Vector Quantizer Performances

IP.com Disclosure Number: IPCOM000101839D
Original Publication Date: 1990-Sep-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 1 page(s) / 39K

Publishing Venue

IBM

Related People

Galand, C: AUTHOR

Abstract

Let us consider a signal x(n) to be vector quantized with a code-book CB containing K code-words of N samples. Each block (x(n);n=1,N) of the signal x(n) is compared with each of the K code-words from the code-book, and results in a best matched code-word CB(k). So, the results of the code-book search is an index k. Then a gain G is computed. The absolute value of the gain is encoded in a b-bit value G' which is transmitted along with its sign S and the index k to the receiver to reconstruct the signal x'(n): x'(n) = S.DECOD(G').CB(k) where DECOD represents a b-bit table containing the decoded values of the absolute value of the vector quantizer gain G, and CB is a copy of the code-book used for the encoding.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 74% of the total text.

Gain Control to Improve Vector Quantizer Performances

       Let us consider a signal x(n) to be vector quantized with
a code-book CB containing K code-words of N samples. Each block
(x(n);n=1,N) of the signal x(n) is compared with each of the K
code-words from the code-book, and results in a best matched
code-word CB(k). So, the results of the code-book search is an index
k. Then a gain G is computed. The absolute value of the gain is
encoded in a b-bit value G' which is transmitted along with its sign
S and the index k to the receiver to reconstruct the signal x'(n):
x'(n) = S.DECOD(G').CB(k)
where DECOD represents a b-bit table containing the decoded values of
the absolute value of the vector quantizer gain G, and CB is a copy
of the code-book used for the encoding.

      One can improve the performances of this vector quantizer as
follows. The amplitude of the signal x(n) is adjusted block by block
by a gain control device which derives a gain g for each block of N
samples. The gain g is, in fact, iteratively computed from its
previous value and from the value of the quantizer gain G' of the
previous block:
                                 b
g = g + TAB(G')   where TAB is a 2 - bit table containing
                   adequate positive and negative increments.
                   g is initialized to 0.

      The input signal x(n) is amplified prior to vector
quantization, and the reconstr...