Linda wrote:
> I have done some real-time measurement i.e. to record the wireless
> multipath channel with a channel sounder persnaphot basis. The data
> was captured by moving the trolley that carried the transmitter while
> the receiver was fixed at one position. My problem now is to choose an
> appropriate method to analysis the data so that I could generalised
> the birth-death of the multipath components(MPCs). After appropriate
> post-processing, I obtained the channel parameters i.e. the delay,T,
> power,P and angle-of-arrival, A of all MPCs at each snaphot. My
> question is, what is the appropriate method that I should use to
> analyse the data so that I could identify the "birth" and "death" of
> these MPCs and the lifespan of these MPCs?

You may want to pick several models and try to find one that fits your
data the best.  Depending on the form of the envelope for each model,
you'll have a different "birth" and "death" time of each component.
The problem is that you can't know the actual envelope of each
component,
a bounce off a tree will be a lot different than a bounce off a moving
car or bus.  

You'll also probably need some kind of iterative analysis - by
subtracting
off the largest components from your data, you can then look for smaller
ones based on what's left.  But by changing the model of the larger
signals - you change amplitude of what you think the remaining
components
are.  Given that you're doing a test in a known environment you should
be
able to find the best match by accounting for the physical positions of
all objects that can give rise to signals.

One method you might be interested in is neural nets.  But you still
need
a good model to start from.

Good luck, that's a mighty hard problem!

Patience, persistence, truth,
Dr. mike
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to