<br><br><div class="gmail_quote">"Alexey Feldgendler" <<a href="mailto:alexey@feldgendler.ru">alexey@feldgendler.ru</a>> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
On Wed, 04 Jun 2008 08:51:54 +0200, Paul-Valentin Borza <<a href="mailto:paulvalentin@borza.ro" target="_blank">paulvalentin@borza.ro</a>> wrote:<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
The classifier is based on multivariate gaussian mixtures. I've<br>
written the discriminant functions and just finished writing the<br>
estimation/training functions (yes, this classifier also needs to<br>
be trained). The classifier is great because it can adapt/learn<br>
new values on the fly while in use. So, when this classifier<br>
classifies enough frames as gestures/motion, it passes these<br>
frames to the recognizer that has a unigram grammar of hidden<br>
Markov models.<br>
</blockquote>
<br>
One thing that worries me is whether continuous recognition is something feasible within Freerunner's battery life and CPU time constraints. Even if the gesture recognizer manages to put the device to sleep when there is no signal and wake up on motion, there will still be a lot of idle processing to do while the user is walking.<br>
</blockquote><div><br>I would think that you just wouldn't have it wake on motion... clicking Aux before starting gesture recognition<br>(if the neo is asleep) seems a reasonable prerequisite.<br><br>
</div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
By the way, it would be great if you guys would have ideas for<br>
gestures. Although creating a new gesture will be trivial, it would be<br>
good for me to know how many I have to create. The demo app will<br>
enable auto-answer when picked up from the table, auto-hang-up and<br>
snooze the alarm when hit. Any more ideas?<br>
</blockquote>
<br>
Is the recognizer able to factor out rotation of the coordinate system? In other words, will it recognize the same gesture if the phone is turned compared to the orientatin used in training, e.g. if the snooze gesture is done from another side?<br>
<br>
Some gesture ideas:<br>
</blockquote><div><snip a bunch of great ideas><br><br>Paul, you might also consider:<br>* holding the moko out & angling the front of it up repeatedly turns up volume<br>* angling front down repeatedly turns down volume<br>
* a set of 5 or 10 standard, easily distinguishable gestures that the user can map to favorite programs<br><br>This is an exciting project; it's great to hear that you've made so much progress already!<br><br>Bobby<br>
<br></div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><br>
-- <br>
Alexey Feldgendler <<a href="mailto:alexey@feldgendler.ru" target="_blank">alexey@feldgendler.ru</a>><br>
[ICQ: 115226275] <a href="http://feldgendler.livejournal.com" target="_blank">http://feldgendler.livejournal.com</a><br>
<br></blockquote></div><br><br clear="all"><br>-- <br>If it doesn't make you smile, you're doing something wrong.