Time delay Measurement via Cross Correlation in DSP board TMS320C6713

Started by hyd198471 October 1, 2010
I am a new developer in DSP board

I am currently working on a project to measure the speed of a small car via
Cross correlation in DSP board.
This small car has embedded two infrared light sensors which send and
receive random signals from the floor. Two random signal are similar except
the time delay when the floor reflects the signals. The sensor signals are
transmitted to the wireless sender and received by the wireless receiver.
The wireless receiver is connected to the DSP board so that the cross
correlation Algorithms can be implemented.
My question is if there is any article related to the time delay estimation
via cross correlation and there is anyone who has experience programming a
cross correlation in the DSP board.
 
Thanks in advance


On Oct 1, 9:33�am, "hyd198471"
<huyuandong@n_o_s_p_a_m.hotmail.com>
wrote:
> I am a new developer in DSP board > > I am currently working on a project to measure the speed of a small car via > Cross correlation in DSP board. > This small car has embedded two infrared light sensors which send and > receive random signals from the floor. Two random signal are similar except > the time delay when the floor reflects the signals. The sensor signals are > transmitted to the wireless sender and received by the wireless receiver. > The wireless receiver is connected to the DSP board so that the cross > correlation Algorithms can be implemented. > My question is if there is any article related to the time delay estimation > via cross correlation and there is anyone who has experience programming a > cross correlation in the DSP board. > > Thanks in advance
One way is to compute the DFT of each signal, conjugate one of the two DFTs, and then multiply on a freq. bin by freq. bin basis the two DFTs together. Then find the argument of each of these results. When plotted as a function of freq. bin number, the result ideally is a line (you may have to unwrap phase) with a slope in direct proportion to the time offset between the two signals. This is a lot easier than peak finding in the cross correlation and also separates freq. offsets from temporal offsets. IHTH, Clay
>On Oct 1, 9:33=A0am, "hyd198471"
<huyuandong@n_o_s_p_a_m.hotmail.com>
>wrote: >> I am a new developer in DSP board >> >> I am currently working on a project to measure the speed of a small car
v=
>ia >> Cross correlation in DSP board. >> This small car has embedded two infrared light sensors which send and >> receive random signals from the floor. Two random signal are similar
exce=
>pt >> the time delay when the floor reflects the signals. The sensor signals
ar=
>e >> transmitted to the wireless sender and received by the wireless
receiver.
>> The wireless receiver is connected to the DSP board so that the cross >> correlation Algorithms can be implemented. >> My question is if there is any article related to the time delay
estimati=
>on >> via cross correlation and there is anyone who has experience programming
=
>a >> cross correlation in the DSP board. >> >> Thanks in advance > > >One way is to compute the DFT of each signal, conjugate one of the two >DFTs, and then multiply on a freq. bin by freq. bin basis the two DFTs >together. Then find the argument of each of these results. When >plotted as a function of freq. bin number, the result ideally is a >line (you may have to unwrap phase) with a slope in direct proportion >to the time offset between the two signals. This is a lot easier than >peak finding in the cross correlation and also separates freq. offsets >from temporal offsets. > >IHTH, >Clay > > >Thanks for your advice >What you mean is simply like following:
Rxy = ifft (fft(x).*conj(fft(y)))
>Rxy should be the cross correlation result of x and y signals. But How I
can find out the time offset in the Rxy cross correlation result
>On Oct 1, 9:33=A0am, "hyd198471"
<huyuandong@n_o_s_p_a_m.hotmail.com>
>wrote: >> I am a new developer in DSP board >> >> I am currently working on a project to measure the speed of a small car
v=
>ia >> Cross correlation in DSP board. >> This small car has embedded two infrared light sensors which send and >> receive random signals from the floor. Two random signal are similar
exce=
>pt >> the time delay when the floor reflects the signals. The sensor signals
ar=
>e >> transmitted to the wireless sender and received by the wireless
receiver.
>> The wireless receiver is connected to the DSP board so that the cross >> correlation Algorithms can be implemented. >> My question is if there is any article related to the time delay
estimati=
>on >> via cross correlation and there is anyone who has experience programming
=
>a >> cross correlation in the DSP board. >> >> Thanks in advance > > >One way is to compute the DFT of each signal, conjugate one of the two >DFTs, and then multiply on a freq. bin by freq. bin basis the two DFTs >together. Then find the argument of each of these results. When >plotted as a function of freq. bin number, the result ideally is a >line (you may have to unwrap phase) with a slope in direct proportion >to the time offset between the two signals. This is a lot easier than >peak finding in the cross correlation and also separates freq. offsets >from temporal offsets. > >IHTH, >Clay > > >Thanks for your advice >What you mean is simply like following:
Rxy = ifft (fft(x).*conj(fft(y)))
>Rxy should be the cross correlation result of x and y signals. But How I
can find out the time offset in the Rxy cross correlation result
On Oct 1, 9:33&#2013266080;am, "hyd198471"
<huyuandong@n_o_s_p_a_m.hotmail.com>
wrote:
> I am a new developer in DSP board > > I am currently working on a project to measure the speed of a small car via > Cross correlation in DSP board. > This small car has embedded two infrared light sensors which send and > receive random signals from the floor. Two random signal are similar except > the time delay when the floor reflects the signals. The sensor signals are > transmitted to the wireless sender and received by the wireless receiver. > The wireless receiver is connected to the DSP board so that the cross > correlation Algorithms can be implemented. > My question is if there is any article related to the time delay estimation > via cross correlation and there is anyone who has experience programming a > cross correlation in the DSP board. > > Thanks in advance
One way is to compute the DFT of each signal, conjugate one of the two DFTs, and then multiply on a freq. bin by freq. bin basis the two DFTs together. Then find the argument of each of these results. When plotted as a function of freq. bin number, the result ideally is a line (you may have to unwrap phase) with a slope in direct proportion to the time offset between the two signals. This is a lot easier than peak finding in the cross correlation and also separates freq. offsets from temporal offsets. IHTH, Clay
I am a new developer in DSP board

I am currently working on a project to measure the speed of a small car via
Cross correlation in DSP board.
This small car has embedded two infrared light sensors which send and
receive random signals from the floor. Two random signal are similar except
the time delay when the floor reflects the signals. The sensor signals are
transmitted to the wireless sender and received by the wireless receiver.
The wireless receiver is connected to the DSP board so that the cross
correlation Algorithms can be implemented.
My question is if there is any article related to the time delay estimation
via cross correlation and there is anyone who has experience programming a
cross correlation in the DSP board.
 
Thanks in advance