(accelerometer-based gestures)

Vasco Névoa vasco.nevoa at
Wed Oct 15 11:50:52 CEST 2008

Hi Paul.
Your tenacity is indeed an example. Congratulations on the fusion of  
your academic and hacking lives! :)
I don't want to burden you even more, but I'd like to remind you of a  
couple of details that may turn out to be important in the long run.
One thing I would like to see come true is the implementation of an  
accelerometer framework that is flexible enough to accommodate all  
kinds of usage, not just gestures, and you are basically sitting on it  
(horay!!!). :)
1 - human gestures;
2 - seismic vibrations (distributed earth quake detection), see [1];
3 - travel "dead reckoning", whether of humans, human-powered  
vehicles, or motorized vehicles;
4 - road comfort level and trail quality level assessment (Z-axis  
vibration for geo map tagging);
5 - morse tapping (for awkward emergencies, you never know...);
6 - anything else you may think of in the future...

My point is, please at least _try_ to include in your design some kind  
of easily extensible framework, possibly allowing dynamic plugins or  
at least compile-in modules, and package your specific gestures work  
as the first modules. And obviously, the ability to support multiple  
kinds of listener clients at the same time. I know it is a lot to ask,  
but I think it is very well worth it.
The future community will thank you very much! :)

Happy hacking!!!  :D



Citando "Paul V. Borza" <paul at>:

> Hi everyone,
> A couple of people asked me whether I will or won't continue my
> project on accelerometer-based gestures.
> My answer was always yes, and to make that clear, I've bought
>, and
> The code has moved from into a GIT
> repository located at
> I'm now in the first year of masters, and I've managed to get two of
> my courses into the accelerometer-based gestures (i.e. to implement
> this as homework).
> From now on, my professors require me to use SOMs (i.e.
> self-organizing maps, a type of neural networks), instead of hidden
> Markov models.
> A couple of you guys asked whether the efficiency and speed can be
> improved using HMMs. Again, the answer is yes, just that I don't have
> time to work on the HMM implementation anymore.
> Just wanted to let you know that from now on, I'll focus exclusively
> on working with the Neo, rather than the Wii to test the gestures; and
> make them smooth and natural.
> Nokia is using SOMs for gesture recognition in mobile phones, so we
> should be on the technology wave as I can tell (still, I'm just one
> guy).
> What I'll focus on in the 0.2 release:
> * use some code from the rotate application that is flying around.
> * keep the current Dbus system for interaction.
> * 10% of CPU (it's now using 20%), and yes it's doable.
> * no GUI, but change the text console UI to be something like 'top',
> and not just printf hundreds of xyz data.
> * reintegrate with matlab-compatible diagrams.
> * will still be in C99 and under LGPL.
> * math formulas that are used in code will have a link to
><page> and the formula will be written
> in LaTex.
> * some of you gave me advices on how to improve the organization of
> the project, will also do that.
> * some dependencies aren't checked, there are too many you say, will  
> be removed.
> * integration with the Dbus FSO communication
> system (I've seen that it grew since I last checked it).
> * implementation of self-organizing maps.
> Bottom line: I'll be trying turn it in a mature project ;)
> You can do interesting things with SOMs, like Nokia was doing:
> detecting when the user climbs down, up or walks, just using an
> accelerometer.
> Thanks,
> Paul
> _______________________________________________
> Openmoko community mailing list
> community at

More information about the community mailing list