really interesting thread: In theory a really rudimental motion detector system could be achived using a REALLY low resolution camera image (even with a 3x3 pixel wide image should be possible to have some informations to use) should be able to recognize when something move from one part of the 'grid' to another, the direction and the speed. Higher the resolution and higher should became the precision (but I suppose will became harder to manage all the extra informations): emulate the shake rock and roll should not be too hard: emulate an acceletometer, well, that sound harder