pi-null-mezon - 2014-04-13

Hello for all, I am very sorry, but I can't provide video guide yet. For a while, try to use application help, and take in mynd three basic rules. For proper aplication work:

1) You shold provide a stationar illumination of the object face (the simplest way to do this is to use desk lamp aimed directly at the face, AC-powered lamps is not very good choice because their luminosity changes with a frequency multiplied by 50 (Russia) or 60 (USA) Hz and you can not see it directly by eye, but probably have seen effect of floating strips on image recorded by digital camcorders with such illumination sources, so such lamps are not very good unless you camera provides suppression of 50/60 Hz interference);

2) You should provide a large face projection on digital image (the measurements accuracy as high as many pixels are processed, so you can come closer to camera or use zoom function);

3) Not all web cams support this. Open "videoCapture" dialog from application and uncheck all auto adjustable functions (seriously, you have to switch all them off), then manually set "sharpness" at low position (because "sharpness" always also amplifies a noise), set "gain" as low as it possible (till face classifier could work), than you should make image brighter by set up "exposure" (high "exposure" better than high "gain" because "gain", once again, always also amplifies noise, but do not set "exposure" too high because it effects on frame rate, for application the best discretezation period will be in range 33-66 ms (30-15 fps successively)).

Ok, check it out, but in addition I should say that for a high accuracy, at a measurement time, your object (his face) should not make large and fast movements on image, but he can close eyes or even blinking (eyes region are not processed), can also talk (but not very fast), or make relatively slow tilts and rotations, all emotions on face will corrupt pulse signal, so try to be serious).

 

Last edit: pi-null-mezon 2014-04-29