Signal coding process and system for implementing said process

Claim 2

A process for coding a signal wherein at least a portion of said signal bandwidth is split into N subbands, the energy contents E(i) of each subband being first determined and the number of quantizing levels q(i) for each subband being then dynamically assigned based on the relative value of E(i) compared to the overall signal energy.

Claim 3

A process for coding a signal including partitioning the signal into fixed length segments, each segment being represented by a block of samples, said samples being then requantized according to the process of claim 1, with each subband energy E(i) being determined for each block of samples, and the quantizing levels q(i) being with B being the overall number of quantizing bits available , and SUM indicating a summing operation.

Claim 4

A process according to claim 2 wherein each block of samples is coded using BCPCM techniques including deriving a characteristic term c(i) from each subband block and deriving the subband quantizing levels

Claim 5

A process according to claim 3 wherein q(i) is adjusted to its nearest integer value y(i).

Claim 6

A process according to claim 4 characterized in that it further includes the following steps: - computing the product of the levels y(i) for the N subbands; - comparing the computed product value to a K value equal to 2**B, and then readjusting the y(i) levels whenever the computed value is different from 2**B.

Claim 7

A process according to claim 4 or 5, further characterized in that it includes optimizing the subband quantizing levels using the following operations: a. compute q(i) for i=1,2,....,N: q(i)=2**(sigma+log c(i)) with sigma = B/N - (1/N) log c(j) ; b. discard bands with q(i)<1 and recompute q(i) for remaining subbands with i being limited to the number of remaining subbands and N is updated accordingly; c. arrange the subbands in some ascending (increasing) priority order (i.e low to high energy and high to low frequency); d. compute candidate integer values y1 (i) and y2 (i) of the levels q(i) for subbands ; e. for each of the remaining bands, set : alpha1 (i) = q(i)/y1 (i) alpha2 (i) = q(i)/y2 (i) ; f. set initial values n=0 ; J(0)=<1> ; e(0))=<0> ; and compute J min (0)=1/PROD alpha1 (i) J max (0)=1/PROD alpha2 (i) PROD indicating a product operation. g. start a recurrent process by increasing the initial value n so that n = n+1 ; h. compute : J min (n)=J min (n-1) alpha1 (n) J max (n)=J max (n-1) alpha2 (n) i. evaluate : e(n,j(n))=minimum with respect to alpha (n) of (alpha(n)**2+e(n-1,J(n-1) for the set <J(n)> = <alpha1, alpha2> * <J(n-1)> the minimization is performed such that at a given point J from the set <J(n)>, e(n,J) should be less then e(n,K) for any K which is greater then J ; hence, if a value J from <J(n)> does not satisfy this condition, it is discarded from the set ; in addition, points which are outside the (J min (n), J max(n)) limit are discarded as well; j. associate with each remaining point in <J(n)> the corresponding value from the set alpha(n) = <alpha1 (n), alpha2 (n)> or the equivalent set y(n) = <y1 (n), y2 (n)> and repeat from step g until n=N (or for all remaining bands) and then proceed further to step k ; k. backtracking : step 0 - select from the final set <J(N)> the optimal value J*(N)>1 which is the nearest to 1 ; step 1 - for J*(N), select its associated values, alpha * (N) from the set <alpha (N)> and y * (N) from the equivalent set <y(N)>; step 2 - compute the origin J*(N-1) of J*(N) in the set <J(N-1)> by : J*(N-1)=J*(N)/alpha * (N) step 3 - Repeat Steps 1-2 for N-1,N-2,...1 ; 1. rearrange the set according to the original sequence.

Claim 8

A process according to claim 6 wherein said candidate integer values are derived through either one of the couples of values : y1 (i) = |q(i)| and y2 (i) = |q(i)| + 1 or y1 (i)=2|0.5(q(i) + 1)| - 1, and y2 (i)=2|0.5(q(i) + 1)|+ 1.

Claim 9

A process according to anyone of claims 2 through 7 wherein the requantized subband samples at each sampling time are packed into one macrosample word W, with . x(i) being an integer value representing the sample on the i-th subband, N being the number of subbands, and y(k) being the quantizing levels of the k-th subband.

Claim 10

A process according to claim 8 wherein said macrosample word : W=x(1)+x(2).Y(1)+......+x(N).Y(N-1) with x(i) being an integer value comprised between 0 and y(i)-1 and representing the requantized sample S(i,j) of the i-th subband at the j-th sampling instant, and Y(k) being equal to the product of levels y(j) for i=1,2,.....,k.

Claim 11

A voice coder comprising filtering means (12) for splitting at least a portion of the voice signal bandwidth into N subbands, and means for coding the signal within each subband using BCPCM techniques wherein a block characteristic c(i) is derived for each subband ; said means for coding comprising BCPCM quantizing means (18) connected to said filtering means, and means (16) for setting said BCPCM quantizing means using said c(i)'s, said setting means being characterized by means for dynamically setting the quantizing levels q(i) for quantizing each subband signal, with:

Claim 12

A voice coder comprising filtering means (12) for splitting at least a portion of the voice signal bandwidth into N subbands, and means for coding the signal within each subband using BCPCM techniques wherein a block characteristic c(i) is derived for each subband ; said means for coding comprising BCPCM quantizing means (18) connected to said filtering means, and means (16) for setting said BCPCM quantizing means using said c(i)'s, said setting means being characterized by means for dynamically setting the quantizing levels q(i) for quantizing each subband signal, with:

Claim 13

A voice coder according to claim 10 characterized in that it includes packing means connected to said BCPCM quantizing means (18) and to said dynamic level setting means, for packing one sample from each subband considered at the same sampling time j, into a macrosample word W:

Claim 14

A voice coder according to claim 11 further including multiplexing means connected to said packing means for multiplexing said W macrosample with the corresponding characteristic terms.

Claim 15

A voice decoder for decoding the multiplexed data according to claim 9 said data including macrosamples W and corresponding characteristic terms c'(i), said voice decoder including : - demultiplexing means for separating each of said W words from its corresponding c'(i) s ; - dynamic level adjusting means sensitive to said c'(i) for deriving the corresponding y(i) therefrom ; - unpacking means connected to said dynamic level adjusting means and to said demultiplexing means for deriving, the subband integer sample values x(i) therefrom; - inverse quantizing means connected to said unpacking means for deriving subband samples S(i,j) ; - an inverse filter bank connected to said inverse quantizing means for deriving voice coded signal samples X(s) therefrom ; and, - digital-to-analog converting means, connected to said inverse filter bank.

Claim 16

A voice decoder according to claim 13, further characterized in that said unpacking means include : - means for setting i = 1 and for deriving x(1) = W modulo y(1) ; - means for deriving a new word value W1 = W - x(1); - means for setting i = 2 and for deriving x(2) = W1 modulo y(2) ; - means for deriving W2 = W1 - x2 ; - means for setting i = 3, and for deriving x(3) = W2 modulo y(2) ; and, so an up to i = N, wherein N is the number of subbands.