Re: Feature proposal: Alternative input system based on low-cost webcam

[Cesar Mauri <cesar crea-si com>]

Hi,

> I think that this raises an important point about inclusion of
> accessibility features: they are often very innovative, and they enhance
> the experiences of average users. For example, the on-screen keyboard
> can be adapted  for use on  touch-screen devices (at least this is my
> understanding from discussions I had with Nohemi Fernandez while we were
> in Montreal) - as long as someone extends the code :)

I agree with that, but we should always keep in mind who is the primary 
target, because sometimes people with disabilities need specific 
features which people without disabilities don't (imagine, for instance, 
a filter for the on-screen keyboard to remove unwanted keystrokes due to 
poor hand control).

> Since the trends right now are moving away from mouse use, it might be
> helpful/interesting to ask ourselves whether or how cameras can be used
> to replace them. Are there situations in which a camera would be more
> useful than a touch screen or track pad? Certainly it can be for certain
> users with disabilities, but we might want to explore other use cases,
> as well.

IMHO (I'm not an expert) we are still in the stone age of the machine 
vision based interaction and we will see great innovations in years to 
come. So, sure, there is much to explore and have fun!

BTW: if someone is interested in playing with this, she might find 
interesting/useful another project called SITPLUS [1] in which we try to 
find other uses of camera based interaction for people with (severe) 
disabilities.

[1] http://sitplus.appctarragona.org

>      > I -think- that for this to work properly we'd need a bunch of things:
>      > first, we need to track not only head movement but also your eyes and
>      > several facial muscles so that we can have accurate tracking and
>     hints
>      > about your
>
>     AFAIK, eViacam developers plans to add support for more facial gestures
>     in the future.

Yes. Some users requested to detect facial gestures [2] and there had 
been some efforts regarding blinking detection [3]. We hope to have time 
(and resources) to add/improve them soon.

[2] 
https://sourceforge.net/tracker/?func=detail&aid=2883828&group_id=248049&atid=1199431

[3] 
http://eviacam.git.sourceforge.net/git/gitweb.cgi?p=eviacam/eviacam;a=shortlog;h=refs/heads/blink_detect

>      > intentions. Well, this requires cameras with resolutions much higher
>      > than VGA, which is the current standard for these. Then, someone
>     needs
>      > to figure out how to track all these elements real-time with little
>      > cpu usage. As it is, we
>
>     With higher resolution cameras the behaviour would be better. But
>     please, read again the feature proposal name "Alternative input based
>     system based on *low-cost* webcam".
>
>     About performance, it is something that was always one of the priorities
>     for eViacam developers, and the reason that the configuration wizard
>     allows you to tweak so many parameters.
>
>      > can't even maintain a Mexican hat over ones head in Cheese without it
>      > lagging 3 seconds behind. And finally for this to work we'de need
>      > pretty good AI to be able to understand what you really want so that
>      > you don't end up sending a draft-mail just because you glanced at
>     that
>      > gorgeous girl that just passed in front of you.
>
>     See my previous comment about performance.

Performance has always been a primary goal. In fact, is not a 
coincidence that we chose C/C++, used multiple threads and wrote custom 
code for the camera capture. And, in my experience, there is no 
noticeable delay in the capture - control loop is the camera is properly 
set up. In fact, I still remember running earlier versions on a PIII 
450Hz laptop without problems.

> I  also think that performance and facial gesture support will most
> likely improve faster if this is included and available to more users
> and developers.

Sure.


Regards,

César