Use a small array of the sensors - one in the center and 3-4 surrounding it. If the heat is detected in the center, it's good, if heat is detected in the side ones, then move the camera until only the center detects it.
Or put the pyro sensor on another scanning stage, and swing it around so you can tell where the person is. You'll also need to limit the field of view of the sensor so that it's about 'head-sized'.
Face tracking is typically done in software, but trying to integrate one in at the last moment is a bad idea.