[Kde-accessibility] simon development

[Mahesh Hegde]

*yes I agree with you in case of classical predator/tracking algorithm.But *
*components used in pertilarly tld apart from tracking,provide necessary
information for recognition.*
* *
*To make sure of these facts I communicated with Author of this
algorithm,who used* *combination of haar waveletts and local binary
patterns, as key features to trackobjects,which we use for recognition.*
*
*
*I gone throuth most of key concepts from his paper.His key components of
model are tracker and detector which agree or decides position of object of
interest.*
*here is complete project details of opentld*
http://info.ee.surrey.ac.uk/Personal/Z.Kalal/tld.html

*this predator is very generic,in this site itself we can see his(autor)
face is also recognised,from set of photos when he shows in front of
cam.one more important fact is we have parameter to vary confidence of
recognition.*
*
*
*regarding depth recognition:this algorithm also provides scaling
information about object(in our case its face).once we know initial
position from cam,then on we can use transformation concepts to determine
distance from cam.even though its very hard to determine depth exactly
because we need stereo information like we need two cameras and it needs
lot of time may be reseach,we get crude aproximation about depth which
definately improve acoustic model.*
*
*
*I am happy if anyone help me to improve or let me know if any flaws in
this approach.*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-accessibility/attachments/20120325/574bc3f8/attachment.html>