after applying a filter -i design- to a real signal, it returns complex signal after ifft

[x,fs]=audioread('some.wav');
n=length(x);
t = ((0:n-1)*(fs/n));

y=fft(x);
f = ((0:n-1)*(fs/n));

y0 = fftshift(y);
f0 = ((-n/2:n/2-1)*(fs/n)); 

noise = find(abs(y0) == max(abs(y0)));
noiseW= [2*pi*f0(noise(1)) 2*pi*f0(noise(2))];

z1=exp(1i*noiseW(1));
z2=exp(1i*noiseW(2));
z=exp(1i*2*pi*f0);
H=(1-z1*(z.^-1)).*(1-z2*(z.^-1));

out0=y0.*(H.');

out=ifftshift(out0);
signal=ifft(out);

audiowrite('output.wav',final,fs);

[Daniel Pollard]

I'm not sure I understand the problem. The FFT (and the IFFT) produce complex outputs by their definition. Could you give some more details? What were you expecting to happen?

[Hamza AKYILDIZ]

some.wav file is a real signal. after i apply the filter H to it, i expect it to be still real after ifft.

Answer 1

[Paul Hoffrichter]

Your imaginary part only is non-zero due to the usual floating point roundoffs and truncations. Fix this using round:

signal=ifft(out);
signalRound = round(signal, 10);