Convolution of Audio Signals

I'm trying to convolve audio signal with impulse response but every time it's giving error as "A & A must be in vector forms"

[sig1, fs] = audioread('mysong1.wav'); % import the song

t = [1:length(sig1)]/fs; 
subplot(3,1,1)
plot(t, sig1) % plot the song

xlabel('t (second)')

ylabel('Relative signal strength')

title('Song')

[sig2, fs] = audioread('StNCh.wav'); % import the song

t1 = [1:length(sig2)]/fs; 

subplot(3, 1, 2) 

plot(t1, sig2) % plot the song

xlabel('t1 (second)')

ylabel('Relative signal strength')

title('Impulse Response')

w = conv(sig1,sig2);
t2 = 0:1:10;
subplot(3,1,3);
plot(t2,w);

xlabel('t2 (second)')

ylabel('Relative signal strength')

title('Convolved Signal')

Please help me to figure out this.

[Walter Roberson]

you read two songs of different lengths with two different sample frequency. You convolve them, which is going to give you a result that one fewer samples than the sum of the two lengths; for example conv(4 samples, 9 samples) would result in 4+9-1 = 12 samples.

What time is each of the output samples if the inputs were different frequency?

You then assume that the result has exactly 11 samples, times 0:10. But it is not.

You need to reconsider this process.

Also I recommend reading the options for conv() and thinking about whether one of them is appropriate.

convolution of signals is effectively using one of the signals as a filter on the other signal, where each additional element of the second signal acts like a further time delay. The second signal is roughly deciding how much echo to add to the first signal, and remember that the adding echo makes a signal longer.

[Shubhra Sinha]

Thank you for your assistance Walter. But what I need to do about the vector error?

[Walter Roberson]

the error now is because your plot has exactly 11 x values, 0 1 2 3 4 5 6 7 8 9 10, but your convolution has many more.

You need to bring the two songs to the same sampling frequency then do the convolution. then construct the x as (0:(length(results)-1))/sampling frequency

[Shubhra Sinha]

Okay, Thank you Walter...

[Ricardo Reyes]

do you have the code fixed up?

Answer 1

[Michael Ndungi]

[sig1, fs] = audioread('mysong1.wav'); % import the song
t = [1:length(sig1)]/fs; 
subplot(3,1,1)
plot(t, sig1) % plot the song
xlabel('t (second)')
ylabel('Relative signal strength')
title('Song')
[sig2, fs] = audioread('StNCh.wav'); % import the song
t1 = [1:length(sig2)]/fs; 
subplot(3, 1, 2) 
plot(t1, sig2) % plot the song
xlabel('t1 (second)')
ylabel('Relative signal strength')
title('Impulse Response')
w = conv(sig1,sig2);
t2 = 0:1:10;
subplot(3,1,3);
plot(t2,w);
xlabel('t2 (second)')
ylabel('Relative signal strength')
title('Convolved Signal')

[Walter Roberson]

This has the same problems I discussed above at https://www.mathworks.com/matlabcentral/answers/520069-convolution-of-audio-signals#comment_833460 and https://www.mathworks.com/matlabcentral/answers/520069-convolution-of-audio-signals#comment_833473

[Nurafiqah Mukhtar]

i cant use this coding.. may i know why?

wave convolution not came out.

[Walter Roberson]

@Nurafiqah Mukhtar

I described some reasons in the comments I posted earlier.