Driver fatigue has getting more attentions from researchers. One of the methods to recognize driver fatigue is by monitoring driver's facial expression. Facial Action Coding System (FACS) is well-known in facial behavior sciences and Action Unit (AU) is used as the measurement. In this work, AU coded facial image is represented by Gabor filters in different orientations and frequencies, which are classified into respective classes using cascaded AdaBoost. The result is validated with ground truth information given in the database. Results show the validation of AU measurement in computer vision method and will be further applied in driver monitoring system to predict the driver's fatigue level.