error in matlab code for snr estimation

Hello all,

I have written following matlab code for snr estimation using the method described in the attached paper. My estimated snr values are not matching with the true snr values. Kindly help.

clc 
clear all;
close all;
% Generate data. 
M = 16; % Alphabet size 
Pd = 10000; % Length of data 
k = log2(M); % Number of bits per symbol
ini_phase=0; % Initial phase
xin = randi([0 1],Pd,1); % Random binary data stream
xin = reshape(xin, length(xin)/4, 4); % Reshape data into binary 4-tuples
xsym = bi2de(xin);                 % Convert to integers
qamsig = qammod(xsym,M,ini_phase,'gray'); % QAM signal 
snr_theory = 0:2:4; % Theoretical value of SNR 
P1=1/4;
P2=3/4;
R(1)=1;
R(2)=3.15;
c2 = 1;
c4 = R(1).^4.*P1+R(2).^4.*P2;
c6 = R(1).^6.*P1+R(2).^6.*P2;
N=2; % Noise variance of 2dB
n = sqrt(N)*[randn(2500,1) + j*randn(2500,1)]; % white gaussian noise
for a=1:length(snr_theory)
    rxsig = qamsig + 10^(-snr_theory(a)/20)*n; % noise addition
    y=abs(rxsig);
    sum_y = 0; 
    sum_y2 = 0;
    sum_y4 = 0;
    sum_y6 = 0;
    for i=1:length(y)
        sum_y=sum_y+y(i);
        sum_y2=sum_y2+y(i)^2;
        sum_y4=sum_y4+y(i)^4;
        sum_y6=sum_y6+y(i)^6;
    end
    M2=sum_y2/length(y);
    M4=sum_y4/length(y);
    M6=sum_y6/length(y);
    m=((M2.*M4)/M6);
    v=[((m.*c6)-c4) ((9.*m.*c4)-c4-4) ((18.*m)-6) ((6.*m)-2)];
    r=roots(v);
    snr_est_M2M4M6 = r(1); % estimation value
end
figure(2);
%plot estimation value 
plot(snr_theory,snr_theory); 
hold on 
plot(snr_theory,snr_est_M2M4M6,'-o'); 
hold off

The graph of estimated snr vs true snr is shown as follows:

Answer 1

[Jan]

Please read: http://matlab.wikia.com/wiki/FAQ#Can_you_program_up_the_algorithm_in_this_article_for_me_and_explain_it_to_me.3F

It is very unlikely that readers in the forum spend the time for reading your paper.