Filter specific frequencies in an audio file

Hello,

First time asking a question on here as I have hit that brick wall now.

I am trying to filer an audio signal that I have, for example I only want to know between 100 and 1000 Hz, instead of the whole thing. I can try a limit of the axis, but of course the data is still there when I further process it.

I have read through many, many, many examples on here and can't seem to figure it out at all. I have looked at bandpass filtering, something to do with butter....

My code uses a FFT to look at the full frequency spectrum, with the audio clip available at http://bit.ly/2RlPl6p . This is just the FFT, I haven't included any filering bits.

File = 'Loud.wav';
[x,fs] = audioread(File);



lpad = 8*length(x);
xdft = fft(x,lpad);
xdft = xdft(1:lpad/2+1);
xdft = xdft/length(x);
xdft(2:end-1) = 2*xdft(2:end-1);
freq = 0:fs/lpad:fs/2;



plot(freq,(abs(xdft)));
hold on
xlabel('Hz')
xlim([0,16000])
ylim([0,3e-3])
ylabel('Amplitude')
hold off

Any help at all would be greatly appreciatied!

(I am not a student doing homework)

P.s.

This is the wav file stats.

 SampleRate: 48000
         TotalSamples: 513984
             Duration: 10.7080
                Title: []
              Comment: []
               Artist: []
              BitRate: 63.1170

Answer 1

[Star Strider]

If you have R2018a or later, use the bandpass (link) function.

If you want to filter it in the frequency domain, see the documentation for the fftfilt (link) function.

[William Osborne]

Hello,

Thank you for the fast response.

I am not overly proficent in matlab, and I have tried to use the bandpass info that you provided over the past week (I follow your answers) , however I have limited idea how to implement it within my code.

[Star Strider]

My pleasure.

The bandpass function has two outputs. The first one is the filtered signal (of the input signal), and the second one (the ‘d’ output) is a digital filter object. Once you provide the necessary frequency, (and if you desire, name-value pair) arguments to the bandpass function, it will provide you with the filtered signal and the filter it designed. Use the second output with the filtfilt function (not filter as the documentation suggests) to filter any other signals you have.

If you want to see the filtering effect, use the fft function to examine your signal before and after filtering. The same fft code will work for both, simply with different time-domain signals. (The sampling frequency and therefore the frequency vector for the fft plot will be the same.)

[William Osborne]

Thank you again.

If I add the bandpass like this:

File = 'Loud.wav';
[x,fs] = audioread(File);

[y,d] = bandpass(x,[100 200],fs)

%??????

lpad = 8*length(x);
xdft = fft(x,lpad);
xdft = xdft(1:lpad/2+1);
xdft = xdft/length(x);
xdft(2:end-1) = 2*xdft(2:end-1);
freq = 0:fs/lpad:fs/2;



plot(freq,(abs(xdft)));
hold on
xlabel('Hz')
xlim([0,16000])
ylim([0,3e-3])
ylabel('Amplitude')
hold off

What do I do next (if the bandpass is even correct), with the filter function? When I have tried running the script I get an error at the bandpass part.

I know you will have far more meaningful things to do, but this is the part I have been stuck at for weeks. Just to get it to work :)

[Star Strider]

My pleasure.

The ‘y’ output of the bandpass function is the filtered signal (here ‘x’), and the ‘d’ output is the digital filter object. (You can request only the ‘d’ output if you want, so you can use it with the filtfilt function — please do not use the filter function with it — with any other signals you want to filter with it.)

What you do with it depends on what you need to do with it. The ‘Loud.wav’ file sounds to me like random noise, and appears to be that when I plot it either in the time domain or frequency domain.

Try this:

File = 'Loud.wav';
[x,fs] = audioread(File);
[y,d] = bandpass(x,[100 200],fs);

N = numel(x);
t = linspace(0, 1, numel(x))/fs;
Fn = fs/2;
FTx = fft(x)/N;
FTy = fft(y)/N;
Fv = linspace(0, 1, fix(N/2)+1)*Fn;
Iv = 1:numel(Fv);

figure
plot(Fv, abs(FTx(Iv))*2, '-b')
hold on
plot(Fv, abs(FTy(Iv))*2, '-r')
hold off
xlim([0 300])
legend('Original','Filtered')

The filter works as expected, and the code runs for me without error.

If you have other signals, I would filter them as:

signal = ...;
filtered_signal - filtfilt(d, signal);

I am not certain what you are doing or want to do, so I cannot currentlly suggest anything oither than I already have.

[William Osborne]

Greatly apprecicated!

If you were wondering what the sound was, it is a river, so is very much a garbled bit of noise.

Thank you

[Star Strider]

As always, my pleasure!

It must be a relatively ‘young’ river with a steep gradient, likely in the mountains. ‘Mature’ rivers in low-gradient topography with meanders are usually silent, or such has been my experience.

[Andrew Moss]

Hi Star Strider, You are amazing :)

I have a recording which contains a chirp signal at 2kHz, I used your code to

filter out any background noise and let only sounds in this (500 - 2500 Hz) frequency pass.

How do I save it back to a wav file now

[x,fs] = audioread(File);
[y,d] = bandpass(x,[500 2500],fs);
N = numel(x);
t = linspace(0, 1, numel(x))/fs;
Fn = fs/2;
FTx = fft(x)/N;
FTy = fft(y)/N;
Fv = linspace(0, 1, fix(N/2)+1)*Fn;
Iv = 1:numel(Fv);
figure
plot(Fv, abs(FTx(Iv))*2, '-b')
hold on
plot(Fv, abs(FTy(Iv))*2, '-r')
hold off
xlim([0 3000])
legend('Original','Filtered')

[Star Strider]

Thank you!

Use the audiowrite (link) function in R2012b and later releases.

[Andrew Moss]

Yes but how do I write only the filtered data?

Is this correct

audiowrite('FileNameGoeshere.wav', FTy, fs)

[Star Strider]

Write the filtered time-domain data as:

audiowrite('FileNameGoeshere.wav', y, fs)

Use the save (link) function to write the Fourier transformed data to a .mat file if you want to. Remember to include ‘Fv’ as well.

[dulanga]

hey star i have a similar question i want to make frequncies higher than 3500 zero and amplify the remianing onces

the unfiltered waw file-

https://www.dropbox.com/s/xg77p5hbi5qr79f/unfiltered_sound.wav1?dl=0

this is my code i still get high pitched sounds...plz help

[y,fs]=audioread('unfiltered_sound.wav1');
f=fft(y);

[ind,~] = find(f>=3500);
b = f;
b(ind) = 0;
c = b*3000;
x=ifft(c);
filename = 'filtered_sound.wav';
audiowrite(filename,c,fs);
clear c fs

[William Osborne]

If you have a new question I Would appreciate if you started a new thread.

Thank you.

[dulanga]

sorry i will

[Kavy Jain]

can somebody plz help me how to extract a given audio signal from set of audio signals