xcorr command makes mistake at 10^-5 order.

function t = xcoor_time_interval(s1, s2, Fs, tit, plot_wanted)

%XCOOR_TİME_İNTERVAL Summary of this function goes here

% Detailed explanation goes here

% Xcoor is put cross correlation of two signals

[c,lags] = xcorr(s1,s2);

% If it wanted xcoor is plotted

if plot_wanted

figure

plot(lags,c);

title(tit);

grid on

end

[~,iLag] = max(abs(c));

t = -(1/Fs)*(lags(iLag));

end

This was my function and main function is:

clc;

clear all;

close all;

addpath('functions');

% Import audio file

filename = '481583__malatestiniccom__gun-3.wav';

% Get audio signal and signals sampling frequancy in Hz

[s1,Fs] = audioread(filename);

s1 = s1(:,1);

s2 = delayseq(s1, -0.00120612644267, Fs);

s3 = delayseq(s1, 0.00102887528457, Fs);

s4 = delayseq(s1, -0.00023602278035, Fs);

s5 = delayseq(s1, 0.00003558800237, Fs);

% Fs = 44100;

% w = 10;

%

% x1 = 0:1/Fs:1;

% x2 = x1 + 0.00120612644267;

% x3 = x1 - 0.00102887528457;

% x4 = x1 + 0.00023602278035;

% x5 = x1 - 0.00003558800237;

%

% s1 = sin(2*pi*x1);

% s2 = sin(2*pi*x2);

% s3 = sin(2*pi*x3);

% s4 = sin(2*pi*x4);

% s5 = sin(2*pi*x5);

plot_wanted = input("Should I plot the figures[true/false]: "); % Determine function should be plot the figure

% if plot_wanted

% figure

% plot(s1);

% hold on

% plot(s2);

% hold on

% plot(s3);

% hold on

% plot(s4);

% hold on

% plot(s5);

% end

% Get time difference by using xcoor refernce is s1 signal

t12 = xcoor_time_interval(s1, s2, Fs, 's1-s2', plot_wanted);

t13 = xcoor_time_interval(s1, s3, Fs, 's1-s3', plot_wanted);

t14 = xcoor_time_interval(s1, s4, Fs, 's1-s4', plot_wanted);

t15 = xcoor_time_interval(s1, s5, Fs, 's1-s5', plot_wanted);

% t12 = vpa(-0.00120612644267,12);

% t13 = vpa(0.00102887528457,12);

% t14 = vpa(-0.00023602278035,12);

% t15 = vpa(0.00003558800237,12);

% t12 = -0.0012;

% t13 = 0.0010;

% t14 = -2.2676e-04;

% t15 = 4.5351e-05;

and like it is shown here the xcorr command makes mistake at the 10^-5 order and it effects my solution much. How can i solve that?

Answer 1

[Pratyush]

Hi @Mustafa Duran,

I understand that 'xcorr' command is not accurate enough at 10^-5 order.

The discrepancy you're experiencing with the `xcorr` command at the 10^-5 order may be due to the limited precision of floating-point arithmetic. To mitigate this issue, instead of using the `xcorr` function, you can try using other cross-correlation methods that might provide better accuracy. One such method is the Fourier Transform-based cross-correlation (`xcorr_fft`). You can also use the Variable Precision Arithmetic (VPA) toolbox in MATLAB to increase the precision of your calculations. See if it helps improve the accuracy of your solution.