Learning involves a rich array of cognitive and effective states.over the past decades research has increasingly highlighted ways in which affective states are central to learning.Learning centered affective states such as engagement and frustration are inextricably linked with the cognitive aspects of learning.Thus understanding and detecting learner affective states has become a become a fundamental research problem.The facial Action Coding System has been widely used to study detailed facial movements for decades.FACS has been widely consumed.This paper presents an automated facial recognition approach to analyzing student facial movements during tutoring and an examination of the extent to which these facial movements corresponds to tutoring outcomes.The result indicate that excellent agreement at the level of presence versus absence of facial movements.Naturalistic video is challenging for computer vision technique.Second the model were constructed to examine whether the intensity and frequency of facial expressions predict tutoring outcomes.CERT produces intensity values for a wide array of FACS facial action unitsthus enabling fine grained analysis of facial expressions .A particularly compelling nonverbal channel is facial expressions which has been intensely studied for decades.