Prev: Why do bandpass filters don't work like notch filters?
Next: Low frequency FFT analysis
From: abdulawal on 28 Apr 2010 08:59
Dear All,

I have used RLS algorithm to predict channel SNR (lets define SNR as x).
SNR is received in frame by frame. The frame duration is 5ms.

Lets define the channel SNR in each frame is xi, for i = 1, 2, 3, ..., t

I used a window of size 4 in RLS algorithm. I want to predict the SNR 2
times in every 4 frame.

At any given frame t, the SNR vector will look like as below.
[xt-3 xt-2 xt-1 xt]

Now after predicting the SNR for next 2 frames, the SNR vector will look
like

[xt-1 xt x't+1 x't+2]

Where x' is predicted SNR.

After prediction there could be some prediction error which I can write
like below

et+1 = |xt+1 - x't+1|
and
et+2 = |xt+2 - x't+2|


Now my question is, can I derive any probability distribution or something
else to quantify the amount of error (e)?

What is the probability of xt+1 in next frame conditioned that xt received
in current frame?

I know the probability distribution of signal is Rayleigh.

Looking forward to hear from you please.

Thanks,
Awal.


 | 
Pages: 1
Prev: Why do bandpass filters don't work like notch filters?
Next: Low frequency FFT analysis