Mickey, I've talked with Daniel recently and told him that I was working on recognizing contexts. For that I'm using self-organizing maps (Nokia is using these for gesture recognition) and I'll try to recognize walking, running, walking up/down stairs etc. Just wanted you to know that I'm continuing my work towards gestures, and walking patterns for the Neo :)
This year I'm working on contexts like said before, but next year I'll try to merge hidden Markov models and self-organizing maps for gesture recognition. These SOM are amazing because you don't need to train them, it just adapts using unsupervised learning. I'm eager to see what will come out of it... Thanks, Paul On Mon, Nov 10, 2008 at 3:51 PM, Michael 'Mickey' Lauer <[EMAIL PROTECTED]> wrote: > Am Monday 10 November 2008 14:14:45 schrieb Atilla Filiz: >> I don't know if there is any work towards this anyway, i think having an >> acc daemon like gpsd would be nice. The daemon can check for blocking in >> acc.meters and reset the hardware if necessary, and serve multiple >> clients(which is not very likely actually). This way maybe we can have more >> quality acc applications, if using them becomes easier and more stable. > > We already have the amazing gesture recognition daemon by P.V.Borza which is > pending integration with the frameworkd. I think most applications will not > want to read the accellerometer data directly, but rather react to signals in > terms of gestures. Those (few) applications that do need direct access can > read from the input device nodes directly. > > -- > :M: > > _______________________________________________ > Openmoko community mailing list > [email protected] > http://lists.openmoko.org/mailman/listinfo/community > -- Paul V. Borza _______________________________________________ Openmoko community mailing list [email protected] http://lists.openmoko.org/mailman/listinfo/community
