Continuous video input at the rate of 60fps is
provided by a camera interfaced with MATLAB. The
individual frame image from the video input is separated and
the location of eyes is detected by making use of the HAAR
cascade method as proposed by Paul Viola [4]. The next step
is deciding whether the eyes are in a closed or an open state.
This is done by obtaining the histogram back projection of
the YUV image of the detected eyes. The probability of skin
presence in each pixel is given by the back projections. The
result is adjusted, if required, by morphological operations
like erode, dilate etc. The closed eyes are detected if there is
a high percentage of skin coloured pixels in the eye region,
otherwise the eyes are open. Then the blink rate for the lie
detection period is calculated and is compared with the
threshold value.