Hand exoskeleton pose monitoring is of great importance in the rehabilitation training of stroke patients to ensure precise robotic trajectory control and provide a patient recovery assessment mechanism. In this paper, a low-cost pose sensor unit based on the principle of photoelectric reflection is proposed to measure the pose of a multi-segment continuous structure in a hand rehabilitation exoskeleton. The sensor unit consists of five photosensitive elements that measure the rotation angle of an arrangement of adjacent segments, each integrated with a sensing element, to estimate the actuator’s motion. An accurate device with a user-friendly interface is then designed for calibration of the sensing elements. The experimental results indicate that the sensitivity exceeds 0.047 V/° for the sensing elements, and hysteresis and repeatability errors are less than 1.1% and 1.8%, respectively. A comparison between the proposed sensor output and the results benchmarked by a VICON motion capture system demonstrates that the sensor can measure the bending angle of the multi-segment structure with a mean error of 3.23 degrees.