A human-machine interface(HMI) using facial expression is expected to realize higher level cooperation between human and machine, because this idea will develop the HMI towards more natural human-machine communication. In this paper, we first introduce an automatic and real-time recognition method of facial expression, and then describe a prototype system developed based on this method. The system captures dynamic facial images from a color CCD camera, extracts feature points on facial parts (eyebrows, eyes and mouth) by real-time image processing, calculates feature values corresponding to the AU of FACS, and recognizes the strength of neutral and six basic expressions for each part of the face (upper, middle and lower) by employing fuzzy estimation.