Amplify and dampen certain frequencies

I would like to amplify lower frequecencies in my audiofile while dampening higher frequecencies. As I see it, I can use a fft and then multiply my signal with a constant to create a dampening/amplifying effect? But this wont create a interpolation between the two states. Now i'm stuck and don't know what to do.

Here is an example of what I want to do in Matlab (0:50 - 1:20): https://youtu.be/YantpouC4Mk?t=50

Any help is appreciated, thanks in advanced!

Answer 1

[Star Strider]

There are several ways to design filters in MATLAB. What you want to do would seem to be most suitable for one of the FIR design functions, such as the Signal Processing Toolbox function firls. They allow you to define the pass-bands and stop-bands as well as the amount of attenuation you want.

Signal processing requires experimentation, so this is necessarily going to be an iterative process until you get the result you want.

My usual filter design procedure for IIR filters is here: How to design a lowpass filter for ocean wave data in Matlab? It’s not as directly applicable to your problem as firls, but contains some ideas that you will probably find helpful, especially if you are relatively new to filter design.

[Sven Thole]

But I don't want to filter out any frequencies. I just want to modify their amplitude. Should I still use this method?

[Star Strider]

Yes.

Please see the documentation for firls and its friends for an illustration on how to design your filter.

[Sven Thole]

Sorry for bothering you.. I have successfully created a filter now! However I been having trouble applying it to my signal:

 [y1, Fs] = audioread('sound.wav');

 Y = fft(y1);

 % Breaking points for frequencies
 o1 = 2 * 1000 / Fs;
 o2 = 2 * 10000 / Fs;

 b = firls(10, [0 o1 o2 1], [4 4 0.25 0.25]);

 % WRONG?
 Y = filter(b, 1, Y);

 y2 = ifft(Y);
 soundsc(y2, Fs);

It gives error: "Error using soundsc. Audio data must be real and floating point." I guess it's because I don't alter the negative frequencies equally with the positive ones. But I can't wrap my head around how to fix it when using a filter.

[Adam]

just use

real( y2 )

if you aren't bothered about the imaginary part of the returned signal anyway.

[Sven Thole]

Tried that. Got some weird results. So was not sure if that the correct way to do it.

[Sven Thole]

Tried using both filter() and filtfilt() functions.

[Star Strider]

No bother at all!

The problem is that you’re filtering the Fourier transform of your signal. Discrete filters operate in the time domain, so your code should be:

[y1, Fs] = audioread('sound.wav');

% Breaking points for frequencies
o1 = 2 * 1000 / Fs;
o2 = 2 * 10000 / Fs;

b = firls(10, [0 o1 o2 1], [4 4 0.25 0.25]);

figure(1)
freqz(b, 1, 2048, Fs)                                   % Filter Bode Plot

y2 = filtfilt(b, 1, y1);                                % The ‘filtfilt’ Function Is Phase Neutral

soundsc(y2, Fs);

I always recommend using the freqz function to be certain that your filter is stable and has the pass-band and stop-band characteristics you want it to have.

The filtfilt function is phase-neutral (maximally flat phase response) so it introduces no phase distortion in the filtered signal, unlike the filter function.

Note — I can’t test your code, but this should work.

[Sven Thole]

This works! Thank you very much, you're the best! :D

[Star Strider]

As always, my pleasure!

Thank you!