Hi! I have performed some propagation channel measurements in frequency domain i.e. I stored the measurement data as complex transfer function. I know I can obtain the channel impulse response by taking the IFFT. My question is, instead of taking IFFT. Can I employ other techniques (usually called super-resolution techniques in radar/wireless communications) such as ESPRIT, EM, SAGE, etc. algorithms to estimate the number of multipaths, their amplitudes and their time-of-arrivals? I have tried to search for papers but I realized that most of them include angle-of-arrivals estimation. For my case, I performed my measurements with single antenna system. So, no angular information can be estimated. Can I still employ these techqiues? Thanks. Linda
Super-resolution technique
Started by ●March 30, 2004
Reply by ●March 30, 20042004-03-30
On Tue, 30 Mar 2004, Linda wrote:> Hi! > > I have performed some propagation channel measurements in frequency > domain i.e. I stored the measurement data as complex transfer > function. I know I can obtain the channel impulse response by taking > the IFFT. > > My question is, instead of taking IFFT. Can I employ other techniques > (usually called super-resolution techniques in radar/wireless > communications) such as ESPRIT, EM, SAGE, etc. algorithms to estimate > the number of multipaths, their amplitudes and their time-of-arrivals? > I have tried to search for papers but I realized that most of them > include angle-of-arrivals estimation. For my case, I performed my > measurements with single antenna system. So, no angular information > can be estimated. Can I still employ these techqiues? > > Thanks. > > Linda >linda, most of the techniques that you mentioned are parametric frequency estimators, meaning that you have to know in advance how many components exist in your signal. there are many ways to do this; for example you can compute the sample covariance matrix of your signal, and compute the singular / eigen values. based on these, you can estimate how many components you have in your signal. for example, if you compute a 5x5 covariance matrix and find 3 large eigenvalues and 2 small ones, you can estimate that you have 3 frequency components, assuming that you have high SNR. this is the basic idea of subspace methods. angle-of-arrivals are estimated by using an antenna array; you compute the _spatial_ covariance matrix, and the angles of arrival correspond to frequency components. in contrast, what i said in the previous paragraphs assume that you compute the sample covariance matrix in the _time_ domain. i think you're confusing things a bit here. i have an introductory document that explains some of this on my website, maybe i should update it at some point. a good reference to frequency estimation would be the statistical signal processing book by munson and hayes. they come with handy matlab codes, too. julius -- The most rigorous proofs will be shown by vigorous handwaving. http://www.mit.edu/~kusuma opinion of author is not necessarily of the institute
Reply by ●March 30, 20042004-03-30
lindah74uk@yahoo.co.uk (Linda) wrote in message news:<124fd8f7.0403300323.1cf4ee46@posting.google.com>...> Hi! > > I have performed some propagation channel measurements in frequency > domain i.e. I stored the measurement data as complex transfer > function. I know I can obtain the channel impulse response by taking > the IFFT. > > My question is, instead of taking IFFT. Can I employ other techniques > (usually called super-resolution techniques in radar/wireless > communications) such as ESPRIT, EM, SAGE, etc. algorithms to estimate > the number of multipaths, their amplitudes and their time-of-arrivals? > I have tried to search for papers but I realized that most of them > include angle-of-arrivals estimation. For my case, I performed my > measurements with single antenna system. So, no angular information > can be estimated. Can I still employ these techqiues?While I do not know very much about telecom, I do know a bit about the types of techniques you consider. It is, generally, not possible to give a definite answer to your question. Everything depends on the application. I have seen some papers where people propose to use MUSIC to reconstruct a time-domain reflection series of a radar echo, thereby presenting a high resolution image of the target. See the paper Kyung-Tae Kim, Dong-Kyu Seo, Hyo-Tae Kim: Efficient radar target recognition using the MUSIC algorithm and invariant features IEEE Transactions on Antennas and Propagation, Vol. 50(3), March 2002, Pages:325 - 337 for an example. I don't know how or why MUSIC works in the radar problem, but I do know that usually, you need some very specific conditions to be met for the method to work. I have written several comments about these things here on comp.dsp in the past, so if you google for something like author:allnor@tele.ntnu.no group:comp.dsp MUSIC at www.groups.google.com, you might find some posts that elaborate on these things. The basic message is that there is no easy way to answer your question. These methods are highly sophisticated, and need to be perfectly matched to the problem to yield meaningful results. This means that only somebody who understand both the methods and the problem can provide useful advice. Rune
Reply by ●March 30, 20042004-03-30
Linda wrote:> Hi! > > I have performed some propagation channel measurements in frequency > domain i.e. I stored the measurement data as complex transfer > function. I know I can obtain the channel impulse response by taking > the IFFT. > > My question is, instead of taking IFFT. Can I employ other techniques > (usually called super-resolution techniques in radar/wireless > communications) such as ESPRIT, EM, SAGE, etc. algorithms to estimate > the number of multipaths, their amplitudes and their time-of-arrivals? >Via angle SR techniques? Yes. However, recall that multipath tends to be correlated highly with the direct signal (i.e. with the signal backscattered from your "target" which also produces the multipath), whilst SR usually requires that the (multiple) sources are uncorrelated.> I have tried to search for papers but I realized that most of them > include angle-of-arrivals estimation. For my case, I performed my > measurements with single antenna system. So, no angular information > can be estimated. Can I still employ these techqiues? >When you say single antenna, I presume you mean a single dish antenna rather than an array of some form (e.g. phased array). Angle SR tends to be performed with array antennas, but you can still do it with a single dish provided it is step-scanning. I can provide a reference if it's relevant. Ta, Carlos
