From Table 1, it can be observed that even though the face
of person 1 undergoes through facial orientation and gesture
variation, the total number of recognized features from the real
time face videos of person 1 always lie in a confined region.
The same thing has been observed for the persons whose image
has not been stored in the system as house member (person 2
and person 3). It has been observed that the variation of
persons causes larger difference in recognizing the features
stored in the system. In order to distinguish between the
detected face of an intruder from the face stored in the system,
a suitable threshold number of recognized features has been
chosen. Alarm and suitable phone autodial should be activated
while a new face is introduced. The threshold for the detected
face to be the same person, whose picture of face has been
stored in the system, has been calculated after changes of facial
orientation and gesture variation and also change of persons.