Audio File (Wav) Filtering

[P3tson]

Hi.
 I am looking for a way to filter the white noise from a wav file. 
There is one input for the wav file, then i added a white noise for it, and i have added a play waveform to the filtered wav file. I have changed the filter settings many times but it is still noisy. Please help me, im noob in labview. 😕

Thanks.
P3tson

Here is the VI and the wav file:
 

[majoris]

What is your signal supposed to be? At what frequencies? 

Below is the Power Spectrum. Time is really Hz and the y-axis is logarithmic. As you can see, there are a few peeks, but they aren't huge over the noise. Spectrum trails off as you go higher in frequency. What are you looking for?

[P3tson]

I just want convert back the orginal signal. Hmm.. my power spectrum before the filltering is empty, but after:

 

it is noisy..

[majoris]

Yeah I don't understand what you're trying to pull out. There's nothing to filter. If you look at the spectrum of the noise, it's flat across frequency: that's the point of white noise. You can't single it out.

If your noise has a particular characteristic (i.e. it's all within a particular bandwidth), you could filter out the noise to recover the signal. Or if you original signal has a particular spectral characteristic, you could do it depending on a variety of factors.

As I indicated before, it looks like most of the signal is between 100 and 500 Hz, so you could try a low pass out to 500 or a band pass from 100 to 500, though that and the smoothing filter seem to do roughly the same thing to my ear. That said, it's never gonna get rid of a noise floor that large!

What are you simulating exactly?

[johnsold]

First, the concept you expressed in the original post appears to be flawed.

Since white noise is uniformly distributed across the spectrum, a filter will not remove it from a signal.  A filter will only reduce the amplitude in the stop band of the filter.

Second, I looked at the spectrum of your audio signal without added noise.  From about 200 Hz to 16 kHz the signal drops about 60-70 dB. In any small segment of the spectrum the local peaks are 20-30 dB above the background.  There is a series of harmonically related peaks spaced 120 Hz and several other peaks without clealry obvious harmonic relationships.  A complicated signal like this is much more difficult to denoise than a well defined signal like a sine or square wave.

The amplitude of your signal peaks at about 0.5 and the average is probably about 0.2-0.3 while your noise has an amplitude of about 3-4. Extracting a signal with a -30 dB signal to noise ratio is moderately difficult if the signal is sinusoidal.  If the decoder does not have access to the original, complicated, signal, it will be nearly impossible to recover it.

Lynn