Forums

Matched filters via cross-correlation

Started by Larry Gadallah March 5, 2016
Hi all:

I am working on a tone receiver that uses cross-correlation to implement a matched filter for the tone. I have frames of samples that I cross-correlate with a frame containing samples of a pure reference tone. My question involves windowing: Should I use a windowing function on each frame? Should the windowing function apply both to the signal frame and the reference frame?

Thanks in advance,
On 06.03.16 02.09, Larry Gadallah wrote:
> Hi all: > > I am working on a tone receiver that uses cross-correlation to implement a matched filter for the tone. I have frames of samples that I cross-correlate with a frame containing samples of a pure reference tone. My question involves windowing: Should I use a windowing function on each frame? Should the windowing function apply both to the signal frame and the reference frame?
If your reference samples are *always* periodic within the FFT block size, i.e. they do not contain frequencies that are not harmonics of FFT length, then you do not need a window function. Maybe you can ensure this. Marcel
Marcel Mueller  <news.5.maazl@spamgourmet.org> wrote:

>On 06.03.16 02.09, Larry Gadallah wrote:
>> I am working on a tone receiver that uses cross-correlation to >> implement a matched filter for the tone. I have frames of samples that I >> cross-correlate with a frame containing samples of a pure reference >> tone. My question involves windowing: Should I use a windowing function >> on each frame? Should the windowing function apply both to the signal >> frame and the reference frame? > >If your reference samples are *always* periodic within the FFT block >size, i.e. they do not contain frequencies that are not harmonics of FFT >length, then you do not need a window function. Maybe you can ensure this.
Odd that the OP said nothing about using an FFT (or even, working in discrete time) yet these things get assumed. Steve
On 06.03.16 21.14, Steve Pope wrote:
> Marcel Mueller <news.5.maazl@spamgourmet.org> wrote: > >> On 06.03.16 02.09, Larry Gadallah wrote: > >>> I am working on a tone receiver that uses cross-correlation to >>> implement a matched filter for the tone. I have frames of samples that I >>> cross-correlate with a frame containing samples of a pure reference >>> tone. My question involves windowing: Should I use a windowing function >>> on each frame? Should the windowing function apply both to the signal >>> frame and the reference frame? >> >> If your reference samples are *always* periodic within the FFT block >> size, i.e. they do not contain frequencies that are not harmonics of FFT >> length, then you do not need a window function. Maybe you can ensure this. > > Odd that the OP said nothing about using an FFT (or even, working > in discrete time) yet these things get assumed.
Doesn't make any difference. Marcel