From: diego@minerva.st.dsi.unimi.it (Diego Montefusco)
Subject: Re: TECH: Kalman filtering
Date: Wed, 3 Feb 93 14:14:28 CET



In <9302021613.AA13323@ghost.dsi.unimi.it>, on Feb 2, you wrote:

> From: Roy_W_Latham@cup.portal.com
> Subject: Re: TECH: Kalman filtering
> Date: Mon,  1 Feb 93 09:28:50 PST
> 
> 
> Kalman filters inherently both predict and remove noise. I suspect that
> filters used by high-end tracker manufacturers are doing predicting as
> well as filtering.

Not to my knowledge. Surely not standard Polhemus Isotrak, maybe the
newer Fastrak. But I don't think: predicting means that you have a
MODEL of user behaviour; a 6D tracker is GENERAL purpose (you can use
it to track head movements, hand, other limbs -for example in medical
applications- even to make a 3D digitizer), so it can't have, IMHO, a
build in predicting filter. Your leg moves quite differently from your
head. Right?

> [description of filtering matrix deleted]

Thank you for the details. As I suspected it can't be done on the same
(hi-end) PC that makes also all the rest (rendering and so on...)
Ah... fundings!

> Incidentally, I think more than half the problem with  tracker lag  is
> really with poorly designed real time systems using the tracker.

Surely it is, but, even with the best designed system you can't do
anything about the HARDWARE delay of the tracker itself. For example
the technical specs for an Isotrak report 43 millisec of delay,
filtered to REMOVE noise.  The new Fastrak claims only for 4 millisec.

Diego