Detecting a noisy signal

[ibbu]

Hi All

Which is the most efficient method to differentiate a noisy signal from a normal one. The application receives signals and it should be able to determine if the signal is noisy or not pragmatically. There is also a threshold value for the spikes in the noise which it should consider.

Thanks for your time in advance...

Regards

Ibbu... 

[pine]

Your question is obscure.

"Noise signal" itself can not provide enough information. For example, when you acquire a mono-tone sound with a microphone, the tone signal or sine signal is a useful signal, the noise is the backgroud noise. But when you acquire ECG signal corrupted by power-line noise, ECG signal is useful signal and the sine signal (power-line noise) is the noise. There are different signal processing methods to seperate "signal" and "noise" based on the different types of "signal" and "noise". For example, if the noise is sine, you can use a simple notch filter to remove it. But if the noise is wide-band and non-stationary, you will need use some advanced signal processing methods like wavelet, Garbor or adaptive filter. 

The best way for you is to put a sample VI, so that I can help you to find a solution. 

[Conseils]

Noise is simply defined as any information that you don't want when it contaminates the information that you do want.

Measurement is performed in the same way as is performed for the signal of interest.

Perhaps the question is 'how  do I seperate the signal of interest from the 'noise' or information that is not interesting'. The basic answer is a filter of some kind. This is where the fun really starts! The most efficient method of removal depends on your signal and the characteristics of the noise.

[Herok]

Hello,

I think your question has been rephrased by Conseils and was answered well by Pine. I feel you have to use some advanced signal processing techniques. Also try using corelation and power matching techniques that may help in your application.

Regards,
Pavan
NI

[ibbu]

Thanks a lot for ur replies...

Will try out the suggestions.

[Chris.Osiecki · Accepted solution]

I agree that an answer depends on what your signal looks like, but it wouldn't have to be that hard to determine if a signal is noisy. If your signal of interest is sinusoidal, you could do an FFT and then threshold the levels other than your target, like a basic signal to noise comparison. That's pretty efficient. If you have something that's changing a lot, like audio or data, again, you could look at the spectrum, but look outside a frequency band, and not just a single frequency. If you have a reference signal, you could subtract the reference signal from the received signal and the result would be your noise and you could measure that. This all gets to determining the quality of the signal, using filtering will help you clean it up, but not tell you how noisy it is to begin with.

Hope that helps.

Chris

[ibbu]

Thanks Chris

[Oli Oli Oxin Free]

Hey chris

How would you do a fft on a audio signal and be able to graph it?

An VI example would be good

Thanks a lot