Forums

How to initialize RLS in practice?

Started by zqchen July 29, 2008
As S. Haykin's Adaptive Filter Theory suggests, we've to choose
different deltas for the initialization of the recursive least squares
algorithm for low, medium and high SNR. Then we've to know the SNR
first. But what if we cannot get the SNR before we initialize RLS? And
in the condition of interference, we've to estimate the SINR first?
thanks.
zqchen <zhiqun.chen@gmail.com> wrote in news:51386e4b-d1a4-4527-8ff4-
76b5e94f7234@j1g2000prb.googlegroups.com:

> As S. Haykin's Adaptive Filter Theory suggests, we've to choose > different deltas for the initialization of the recursive least squares > algorithm for low, medium and high SNR. Then we've to know the SNR > first. But what if we cannot get the SNR before we initialize RLS? And > in the condition of interference, we've to estimate the SINR first? > thanks. >
From one point of view, theres design, and then there's implementation. In the design stage, you learn what your SNR is, and you use this information to pin down all your parameters, deltas, and such. If you really have a situation in which SNR is changing drastically enough such that working with the wrong delta for your SNR makes that big of a deal, perhaps your approach needs reconsideration. -- Scott Reverse name to reply
On Jul 30, 2:49 am, zqchen <zhiqun.c...@gmail.com> wrote:
> As S. Haykin's Adaptive Filter Theory suggests, we've to choose > different deltas for the initialization of the recursive least squares > algorithm for low, medium and high SNR. Then we've to know the SNR > first. But what if we cannot get the SNR before we initialize RLS? And > in the condition of interference, we've to estimate the SINR first? > thanks.
Shouldn't make much difference. a large delta speeds things up a bit. K.