CONCURRENT PROCESS FOR BLIND DECONVOLUTION OF DIGITAL SIGNALS

• US 20070189374A1
• Filed: 01/12/2007
• Published: 08/16/2007
• Est. Priority Date: 03/02/2001
• Status: Active Grant

• Current Assignee: Unlao Braslleira De Educacao E Assistencia
• Original Assignee: Unlao Braslleira De Educacao E Assistencia
• Inventors: Arantes, Dalton, Cristina Felippetto De Castro, Maria, Comparsi De Castro, Fernando
• Granted Patent: US 7,433,400 B2
• Time in Patent Office: Days
• US Class Current: 375/232
• CPC Class Codes: H04L 25/03006 Arrangements for removing intersymbol interference
• CPC Class Codes: H04L 25/03019 adaptive, i.e. capable of adjustment during data reception
• CPC Class Codes: H04L 25/0307 using blind adaptation

Abstract

A process and system for eliminating intersymbol interference in digital signals, which is caused by the dispersive effect of any practical transmission channel.

Claims

1. A system for blind deconvolution of digital signals, the system having the following form:

in which;

A) Output signal y from the system is a sum of output signals of two filters, V and w;

B) Vector V=[V₀ V₁ . . . V<sub>L−

1</sub>]^T is adjusted for any gradient based algorithm that minimizes a cost function J_D, that measures dispersion,

C) Vector W=[W₀ W₁ . . . W<sub>L−

1</sub>]^T is adjusted for any gradient based algorithm that minimize a cost function J_Q, or any other equivalent function that measures the distance of the output y from the nearest digital alphabet symbol, denominated as Q{y}, where the operator Q{.} represents quantization of the symbols of the digital alphabet A;

D) Any block z<sup>−

1</sup> introduces a delay of one sample;

E) E{.} is the operator that restores a statistical median o f the argument;

F) Operator {.}^T results in the transposition of a vector/matrix argument;

G) Operator |.| returns the Euclidian norm of the vector/matrix argument; and

H) A connection between J_D and J_Q is obtained by means of a non linear function that inhibits the process of J_Q when the minimization process of J_D does not simultaneously minimize J_Q.

2. A system for blind deconvolution of digital signals, the system having the following form:

in which;

A) Output signal y from the system is a sum of output signals of two filters, V and W;

B) Vector V=[V₀ V₁ . . . V<sub>L−

1</sub>]^T is adjusted for any gradient based algorithm that minimizes a cost function J_D, that measures dispersion,

C) Vector W=[W₀ W₁ . . . W<sub>L−

1</sub>]^T is adjusted for any gradient based algorithm that minimize a cost function J_Q, or any other equivalent function that measures the distance of the output signal y from the nearest digital alphabet symbol, denominated as Q{y}, where the operator Q{.} represents quantization of the symbols of the digital alphabet A;

D) Any block z<sup>−

1</sup> introduces a delay of one sample;

E) E{.} is the operator that restores a statistical median of the argument;

F) Operator {.}^T results in the transposition of a vector/matrix argument;

G) Operator |.| returns the Euclidian norm of the vector/matrix argument; and

H) A connection between J_D and J_Q is obtained by means of a non linear function that inhibits the process of J_Q when the minimization process of J_D does not simultaneously minimize J_Q,

wherein the system is carried out by a concurrent process for blind deconvolution of digital signals, comprising the steps of;

a) Initializing the vectors W and V;

W=0+j0 and

$V_k=\{\begin{array}{c}0+j<br><br>\text{\hspace{1em}<br><br>}<br><br>0,k=0,1,\mathrm{\dots <br><br>}<br><br>\text{\hspace{1em}<br><br>}<br><br>L-1,k\ne <br><br>\mathrm{\xi <br><br>}\\ 1+j<br><br>\text{\hspace{1em}<br><br>}<br><br>0,k=\mathrm{\xi <br><br>}\end{array}$

 

where j=√

{square root over (−

1)}, L is the dimension of the vectors V and W, and ξ

, 0<

ξ

<

L−

1, is the index of the vector V to be initialized with the value 1+j0;

b) Initializing the indexer of samples obtained for fractional sampling of the channel;

i=1;

c) Initializing the regression indexer of the channel;

n=0;

d) Obtaining the n regressor of the transmission channel r(n);

r_k(n)=u(L−

1−

k+i), k=0, 1, . . . , L−

1, where the u is the sequence of samples received for the fractionary sampling T/2 of the transmission channel with i=1, 3, . . . N_a−

1 changing as n=0, 1, . . . , N_r−

1, N_a is the total number of samples that will be obtained for fractionary sampling of the transmission channel,

$N_r=\left[\frac{N_a-L-1}{2}\right]+1$

 

is the total number of regressors to be obtained of the transmission channel and T is the time gap between the symbols generated in the transmitter, └

.┘

is the operator that results in the nearest whole number smaller than the argument;

e) Obtaining the output of the system in the instant n;

y(n)=W^T(n)·

r(n)+V^T(n)·

r(n);

f) Updating the vector V;

V(n+1)=V(n)+η

_v·

y(n)(γ

−

|y(n)|²)·

r*(n), where η

_v is the adaptation step of the vector V, 0<

η

_v<

<

1.0;

g) updating the vector W(n+1)=W(n)+η

_w[1−

D_Q(n)][Q{y(n)}−

y(n)]r*(n), where η

_w is the adaptation step of the vector W, 0<

η

_w<

<

1.0, and

$D_Q=\left(n\right)=\{\begin{array}{c}0,Q<br><br>\left\{y<br><br>\left(n\right)\right\}=Q<br><br><br><br>\stackrel{\_}{y}<br><br>\left(n\right)<br><br>\\ 1,Q<br><br>\left\{y<br><br>\left(n\right)\right\}\ne <br><br>Q<br><br><br><br>\stackrel{\_}{y}<br><br>\left(n\right)<br><br>\end{array}$

 

controls the non-linear function and {tilde over (y)}(n)=V^T(n+1)·

r(n)+W^T(n)·

r(n);

h) Increasing indexer l to i+2 and n to n+1; and

i) Testing end of loop and if L+i>

N_a, then ending the process and in any other case repeating steps d) to i).