FFT Filter

[mereaux]

Hello everybody,

I am trying to look at the FFT of a sound signal. I used for it scilab.
The problem I have is that there is a lot of noise,
and I do not know how to remove it.
What kind of filter should I use?

Please answer simply, I am not used to signal treatment analyzis!
thank you

mereaux

 

[sreid]

Consider two simple options first.

Moving Average Filter (MA Filter). Continuously average the N latest samples.

Oversample and Average. Oversample at Nf and average every N samples.

Both of these techniques reduce the noise by a factor of the square root of N. They are also optimal in the sense that one gets the greatest noise reduction for a minimum decrease in the signal bandwidth.

 

[nbucska]

If the signal is periodic, averaging is the best way.
(the "bestest" :use PLL to sync the sampling )
If the signal is band-limited, use band/low-pass filter.
If the noise is narrowband, use notch-filter.

If the noise is external, use more directional mike.

and read the FAQ:







<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[jsolar]

A filter will not improve your view of the FFT output. You will only reduce the amplitude of the pickets. If the sound signal has the same frequency components that your noise has you can not use frequency based filtering very well. You will be able to improve your Signal to Noise Ratio (SNR), but you will add distortion. So, you need to determine the characteristics of the noise; are they temporal or can you define a frequency Power Spectral Desnity (PSD). You then need to know the same thing for the signal. To maximize your SNR you woudl do a pre-whitening filter. However, you still will not change what you see in the frequency domain. Can you describe your requirements? What is your objective?
John Solar