Denoising and envelop detection

Hi, I am working on radar processing and I am new to this study. I have 1D
time domain data and I used windowing technique and 'fft'ed the data and
plotted the data using imagesc and it is a spectrogram. My question is to
denoise the spectrogram and to extract the data from the envelope obtained
in the spectrogram. I am working this study in Matlab. Can any one suggest
me denoising algorithm and envelope detection.

Thank you.

On Feb 23, 8:22&#4294967295;am, "wolverine_w2000" <jeetu...@gmail.com> wrote:
> Hi, I am working on radar processing and I am new to this study. I have 1D
> time domain data and I used windowing technique and 'fft'ed the data and
> plotted the data using imagesc and it is a spectrogram. My question is to
> denoise the spectrogram and to extract the data from the envelope obtained
> in the spectrogram. I am working this study in Matlab. Can any one suggest
> me denoising algorithm and envelope detection.

Noise is separated from signal by knowing something that is different
between the two.  You use that difference to separate them.  What do
you know about your signal and the noise that you could use to
separate them?  What can you change about the way you collect the
data?

Rick

On Feb 23, 8:22&#4294967295;am, "wolverine_w2000" <jeetu...@gmail.com> wrote:
> Hi, I am working on radar processing and I am new to this study. I have 1D
> time domain data and I used windowing technique and 'fft'ed the data and
> plotted the data using imagesc and it is a spectrogram. My question is to
> denoise the spectrogram and to extract the data from the envelope obtained
> in the spectrogram. I am working this study in Matlab. Can any one suggest
> me denoising algorithm and envelope detection.
>
> Thank you.

If this is a pulsed time domain signal, I'm confused as to why you'd
want to envelope detect it in the frequency domain. You will lose time
resolution on the pulse edges, which is typically important. You could
perform frequency domain processing to "denoise" the signal and then
return to time domain prior to envelope detection. As mentioned
before, any algorithm will exploit some difference between signal and
noise. For example, you might know that the signal spectrum is
symmetric and restore that property by adjusting FFT bins on opposite
sides of the carrier frequency. Or, you might know that signal is
absent in certain intervals and create a noise spectrum to be divided
out during intervals when signal is present.

In the time domain, you might consider applying a median filter
followed by a matched filter.

John