Create a simulation scenario

Hi all, I have cobbled up a plan. I want to know if it makes sense.

First part (1) Divide a room (or any closed space) into several regions, R. (2) Place a UWB receiver(rx) node at a fixed position somewhere outside R. (3) Place a UWB transmitter(tx) node inside one of the regions. Now send some pulse from the tx to the rx. Record channel impulse response (CIR) of the tx-rx pair. Take some (say, n) measurements from within that area, but placing the tx at a different point each time. So we have a bunch of CIRs. (4) Store the CIRs in a database. Since I want to use Matlab, let's say I have thus created a vector with n entries. I assume that the index of the vector tells me from which region the CIR was recorded.

I am using a CIR generator based on ray tracing to simulate the CIR received at the rx.

Second Part

In the second part,I want to send some signal just like before, but this time I pretend that I don't know from which region the signal is being sent.

So I take that unknown signal and correlate with all the entries of the previously stored database.If I get a high correlation value at some index of the database, the chances are that they both originate from the same region. Thus I have created a region identification system!

To create the unknown signal, all I do is generate the CIR with the same generator,and add some AWGN using Matlab, like this: noisy_h = (clean_h, SNR);

Now, I correlate this noisy_h with all the entries of the database. I arbitrarily add 3dB, 6db, 9dB,... etc and create noisy data for that particular SNR. In my experiment, as the SNR goes up, the correlation values also go up.I think it is natural.

For a particular SNR, I repeat the correlation process, say, 1000 times (with a loop) and if I get correlation values more than T(some threshold), I assume that as hit. Say, out of 1000 times, I got 890 hits. Can I say that my probability of detection is 0.89?

If not, how can I simulate probability of detection in this case?

So that was it. Does it make sense?

Thank you for taking time to read this long post.

Cheers,

PS: When creating a noisy signal, can I add SNR arbitrarily? Is it ok if I find out variance of the clean_h and add that directly as described above?