Lets say I have a system where I have a small sample of a signal with no noise $\hat{x}(t)$ and a lot of a similar signal with noise $y(t) = \hat{x}(t) + n(t)$, and from $\hat{x}(t)$ I want to create a matched filter to 'find' $\hat{x}(t)$ in $y(t)$

In theory I would go about doing this by first finding $\hat{X}^*(k) = F^*_k[\hat{x}(t)]$

So far, no issue, but my signal $\hat{x}(t)$ is 10 minutes long sampled at 100Hz and non-stationary, so when I have previously been analysing the power in this signal i have been using Welch's method to find the power in small increments of $\hat{x(t)}$ which can be assumed to be stationary and then averaging which helps to remove the noise.

I would like to do something similar to try and have a more accurate transfer function, unfortunately I think I am correct in saying that you can't average small, Fourier transformed, increments in the same way because of the interference caused by the phase difference.

So if I wanted to match $y(t)$ with $\hat{x(t)}$ would I simply take one small sample of $\hat{x(t)}$ and use that to find a transfer function, or are there more clever ways?


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.