We describe a real-time stereo camera system for tracking a human hand from a sequence of images and measuring the three dimensional trajectories of the hand movement. We incorporate three kinds of constraints into our tracking technique to put restrictions on the search area of targets and their relative positions in each image. The restrictions on the search area are imposed by continuity of the hand locations, use of skin color segmentation, and epipolar constraint between two views. Thus, we can reduce the computational complexity and obtain accurate three-dimensional trajectories. This paper presents a stereo tracking technique and experimental results to show the performance of the proposed method.