I have some pattern signal. This pattern signal will be played in noisy room. Then it will be recorded and processed. I need to show that my processing algorithm improves the signal.

My idea is to cross correlate pattern with recorded signal and improved signal and compare these two numbers. If number is bigger then algorithm improved the signal if it is smaller then it decreased. This method is easy but numbers I get are not possible to interpret anyhow but to say more is better.

I would like to have some way of measuring SNR on both signals. Problem is that when I have pattern I don't know how to separate my pattern signal from noise to calculate SNR. I was thinking about just substracting recorder - pattern = noise but I'm not sure about this because my pattern will be modified by room and microphone characteristics and of course it can have different volume so I should weight pattern samples by some constant.

Are there some standard robust way of doing such thing?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.