A method and system for monitoring a driver is disclosed. Firstly, an inner cabin image of a vehicle's cabin is continuously captured. Next, a face detection of a driver for the inner cabin image is performed to obtain a face detection result. Next, the inner cabin image and the face detection result corresponding to the inner cabin image are stored. When the system observes that either the face of the driver is not detected or the face is not shown within a reasonable face region of at least two continuous inner cabin images according to each face detection result, the system outputs a warning signal to alert the driver. The invention determines a driving state of a driver in a cabin with a method of analyzing a position of a face or a head and using an image extractor having large capture range and a low resolution.