Delete Frequency in FFT File and create new WAV-File

Hello supporters,

I have a 0,25 seconds audio file in wav format where i want to delete some frequencies.

1) First i make a FFT to get the spectrum

2) Then i put in some Frequency i want to delete and a tolerance

3) When i've deleted the Frequency i make a IFFT to get the origin signal without the deleted frequency

I'm little bit confused from my result i got after the IFFT. I've deleted all frequency above 400Hz and in my IFFT x(t) are higher frequencys than that?

Did i something wrong in the the shifting from the fft because its symmetric?

I've also attached the WAV File

Many thanks in advance for help!! BR Mathias

clear all;
close all;


%Wav File einlesen und die Daten in data und Abtastrate in Fs
[data,Fs]=audioread('Mono_120A_v20_02sec.wav');
%Ermittlung der Arraygröße
[nSamples,nChannels]=size(data);
%Länge der Audiodatei ermitteln
waveFileLength=nSamples/Fs;


t=[0:length(data)-1] / Fs;

%Signal darstellen
subplot(3,2,1)
plot(t,data)

title('Audio')
xlabel('Zeit')
ylabel('X(t)')


%FFT des Signal

y_fft = abs(fft(data));               %Retain Magnitude
y_fft = y_fft(1:nSamples/2);       %Discard Half of Points
f = Fs*(0:nSamples/2-1)/nSamples;  %Prepare freq data for plot


%Plot Sound File in Frequency Domain
subplot(3,2,2)
plot(f, y_fft)
grid on
xlim([0 20000])
% ylim([0 200])
xlabel('Frequency (Hz)')
ylabel('Amplitude')
title('Freqenzbereich 70A')


yy=fft(data);
yyy=fftshift(yy);
f=Fs.*(-nSamples/2:nSamples/2-1)/nSamples;
subplot(3,2,3)
plot(f,(yyy.*conj(yyy)/(nSamples*nSamples)));
title('FFT')

%%%%%%%%%%%%%%%%%%%

Fdelete = 15000; % Diese Frequenz soll gelöscht werden


tolerance = 14600; % Toleranz um die Frequenz
idx_p = find(f>Fdelete-tolerance & f<Fdelete+tolerance); % Löschen der Frequenzen
idx_n = find(f>(Fs-Fdelete)-tolerance & f<(Fs-Fdelete));
idx = [idx_p idx_n];
yyy(idx) = 0;

Fdelete = -15000;

tol = 14600; % Toleranz um die Frequenz
idx_p = find(f>Fdelete-tolerance & f<Fdelete+tolerance); % Löschen der Frequenzen
idx_n = find(f>(Fs-Fdelete)-tolerance & f<(Fs-Fdelete));
idx = [idx_p idx_n];
yyy(idx) = 0;
subplot(3, 2, 4)
% plot(f, yyy); title('After');
plot(f,(yyy.*conj(yyy)/(nSamples*nSamples)));
title('After')

%%%%%%%%%%%%%%%%%%%

subplot(3, 2, 6)
plot(t,data)

title('origin x(t)')
xlabel('Zeit')
ylabel('X(t)')


%%%%%%%%%%%%%%%%%%%

subplot(3, 2, 5)
y1=ifft(yyy,'symmetric');

plot(t,y1),grid on;
title('IFFT x(t)')

[Ignacio Jordan]

hello. did you get the answer?

Answer 1

[Mathieu NOE]

Hello

I would rather do a notch filter on the time data

if multiple frequencies to eliminate, do a loop

notch filter - see example of digital filters attached