Received: from watserv1.uwaterloo.ca (watserv1.waterloo.edu) by karazm.math.UH.EDU with SMTP id AA09615 (5.65c/IDA-1.4.4 for ); Tue, 15 Oct 1991 13:51:31 -0500 Received: by watserv1.uwaterloo.ca id ; Tue, 15 Oct 91 14:47:19 -0400 Date: Tue, 15 Oct 91 14:47:19 -0400 From: Dave Stampe-Psy+Eng Message-Id: <9110151847.AA05571@watserv1.uwaterloo.ca> To: glove-list@karazm.math.uh.edu, schildba@spot.Colorado.EDU Subject: Re: LPC vs. polynomial predicting I think one of the problems with ANY predictive filtering of the poer glove is going to be the sparseness of the data (that is, the wide spacing of the samples). These methods work best with finer data. BTW, an idea about speeding the glove's readout: from my fooling around, the glove's computer reads the receivers after pulsing the two transmitters (20 mS each) then reads the fingers with an RC circuit and comparator for each finger. This takes the remaining 35 mS of the 75 mS glove cycle: in fact, when you bend the fingers, you can *hear* the transmitter's click rate change, showing that the finger read cycle gets shorter. SO: why not "hardwire" the finger sensor RC circuits low (or high, I forget which) and cut 35 mS off the sample period? An external A/D converter could read the finger positions with at least 4 bits of precision. Of course, there are obvious disadvantages here, but it might push the sample rate up to 25 Hz, which is much better than the current 13.3 Hz -Dave Stampe