How can i calculate the signal to noise ratio of a signal without knowing signal or noise power?

I have a noisy signal and i know the sampling frequency, amplitude and length. But i don't have the signal or noise power. Is there any way to calculate SNR?

Answer 1

[Greg Heath]

If the noise is additive with the majority of the noise power outside of the frequency range of the signal, the signal can be estimated by filtering out most of the noise. The noise is then estimated by subtracting the signal estimate from the noisy measurement. The ratio of the power of the two estimates yields an estimate of the SNR.

If the noise is multiplicative, the logarithm of the square of the measurement is used.

Hope this helps.

Thank you for formally accepting my answer.

Greg

[Jan]

@Greg: Then we have to assume, that the OP knows the frequency range of the signal. But Minhaj explains, that only "sampling frequency, amplitude and length" is known. Do I miss a point?

[Greg Heath]

Usually a quick look at the FFT will indicate whether the approach is feasible.

...A common approach in radar and sonar data.

Answer 2

[Jan]

No, of course not. To determine or even estimate the Signal to Noise ratio, you have to distinguish the Signal and the Noise. If you do not have further information, this is impossible. If you know, that the signal is e.g. a clean sine wave, such a distinction would be possible.

Answer 3

[Arsalan]

Usually in Experimental work. people use the spectrum of the signal to distinguish the SNR. Thus I would assume if you change your data such that they signify power and take a FFT of the data (given that the data you have or in time domain). You'll chose a reference bandwidth and get the difference between your noise floor power and the data power (in this FFTd version of data), the resultant value will be your SNR within a given bandwidth.

Arsalan

[Kiranraddi morab]

please can you elaborate your answer with example??