In order to develop an algorithm of distraction warning for heavy-duty truck drivers, a subjective evaluation of forward careless state which should have to be warned in case of distraction, was conducted with a various driving behavior clipped scenes. Then, an estimation algorithm for forward careless state to be warned was devised, based on the face orientation/position, recognized by Driver Monitoring System (DMS). The algorithm was implemented on the DMS embedded in the A pillar of the cabin, and evaluated on an actual traffic road. As a result, it was confirmed that this newly developed algorithm consists a good capability of detecting various types of distraction properly.