To all the members of this community, iam doing my project on ultrasonic signal processing. i need to design a matched filter,to remove the white noise from the received signal(it is the signal received by ultrasonic receiver sensor). as iam having very little knowledge about signal processing i dont know how to design a matched filter practically for a given frequency around 40KLHz.The only thing that i know about matched filter is that it performs mathematical convolution function. CAn anybody help me to do that?
How to design a matched filter?
Started by ●March 25, 2009
Reply by ●March 25, 20092009-03-25
On 25 Mar, 11:59, "vimmi" <vimal.veeren...@gmail.com> wrote:> To all the members of this community, > iam doing my project on ultrasonic signal processing. > i need to design a matched filter,to remove the white noise from the > received signal(it is the signal received by ultrasonic receiver sensor). > as iam having very little knowledge about signal processing i dont know > how to design a matched filter practically for a given frequency around > 40KLHz.The only thing that i know about matched filter is that it performs > mathematical convolution function.No, it doesn't. First of all, read up on DSP. After having done that, you need to find out exactly what signal you look for. A 'matched filter' is exactly that - matched - so you can't use any filter with any signal. In other words, you need a lot of information: - The nominal pulse as emitted by the transmitter signal generator - Any distorsions of that pulse through the transducer and into the medium under measurement - Any distorsions of the recieved pulse through the reciever Only when you have a clear idea of these factors do you know what signal your matched filter should look for. In addition, you might have to compensate for non-linear effects, if this is high-power systems. Rune
Reply by ●March 25, 20092009-03-25
Rune Allnor wrote:> On 25 Mar, 11:59, "vimmi" <vimal.veeren...@gmail.com> wrote:...>> The only thing that i know about matched filter is that it performs >> mathematical convolution function. > > No, it doesn't. First of all, read up on DSP.Don't all signal filters (distinguished from coffee filters) convolve in time? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●March 25, 20092009-03-25
On 25 Mar, 14:47, Jerry Avins <j...@ieee.org> wrote:> Rune Allnor wrote: > > On 25 Mar, 11:59, "vimmi" <vimal.veeren...@gmail.com> wrote: > > � �... > > >> The only thing that i know about matched filter is that it performs > >> mathematical convolution function. > > > No, it doesn't. First of all, read up on DSP. > > Don't all signal filters (distinguished from coffee filters) convolve in > time?OK, let's ask from the other direction: If I state a FIR filter, say, b = [1 2 3] or an IIR filter, a = [1 1.8 0.8] b = [1 0 1] I just see lists of numbers. Where are the convolutions? Rune
Reply by ●March 25, 20092009-03-25
On Mar 25, 6:59�am, "vimmi" <vimal.veeren...@gmail.com> wrote:> To all the members of this community, > iam doing my project on ultrasonic signal processing. > i need to design a matched filter,to remove the white noise from the > received signal(it is the signal received by ultrasonic receiver sensor). > as iam having very little knowledge about signal processing i dont know > how to design a matched filter practically for a given frequency around > 40KLHz.The only thing that i know about matched filter is that it performs > mathematical convolution function. > �CAn anybody help me to do that?What you're looking for is probably a Wiener filter, not a matched filter. Matched filter is for decoding of a communication signal. Since they are both a "filter," they are implemented by convolution. Now, are you looking for help on: 1) Designing a Wiener filter, or 2) Implementing one, given that you have a design, or 3) Both? To design a Wiener filter, you need to assume that your signal of interest is stationary, and you need to estimate or derive its spectrum. If you do a google search, there are many explanations of how to do a convolution in software or hardware. Julius