Introduction. Biomechanists have been trying to obtain integrated and accurate human motion data. However, it is not so easy, because some have limitation of accuracy, some have limitation of the observation area, and some are expensive. For example, motion capturing can obtain whole body motion data, but needs space, is expensive, but only surface motion could be obtained. So is not so sensitive for the bone rotation. Sensors of pressure, acceleration, and so forth are less expensive and less hard to use, but the data are limited. 2D–3D shape matching such as Jointtrack can describe bone motion including rotation, but the detectable area is limited by the size of flat panel fluoroscopy. In this study, we have combined multiple joint motion analysis by Jointtrack and reconstructed full lower extremities' motion. Method. Pelvis, bilateral femurs, and bilateral tibiae geometries were obtained from CAT scan using Mimics®. Gait motion fluoroscopy was done on a treadmill around hip joints and knee joints (Fig.1). On each heal thin film switch was attached and connected to electrically driven metal flag which can be recorded in fluoroscopic images on heal strikes. Images of five gait cycles were taken with 15Hz and every image was sorted by the percentage of gait cycle, and then processed by Jointtrack. Centre of femoral head of observing side was defined as our origin. Using treadmill, the walking direction could be uniquely defined. From the femoral 3D displacement and rotation, knee position can be calculated. The same procedures were done for the knee assessment, mutual coordinate of hip centre from the knee can be calculated. All of them are sorted by the percentage of the gait cycle too. Combining data from hip and knee, complete lower extremities' motion could be described. Regression analyses of x, y, z coordinates of femurs from hip and from knee were done to evaluate the accuracy. Motion capture of floor gait and treadmill gait were done to evaluate the difference. (Fig.2). Materials. Five arthritic hip patients and one normal control were assessed in hip and knee by Jointtrack. Normal Control was assessed by motion capture too. Results. The motions of lower extremities were descrobbed as the translations of coordinates of hip, knee, and ankle joints (Fig.3). The coefficients of determination were calculated for each patient. Except for one patient, good regression was obtained between hip assessment and knee assessment. In motion capture, the difference of the gait speed on the floor and on the treadmill effected to the gait pattern. Discussion. Motion capture can assess multiple joint motions in natural state, but they are only skin motion. Jointtrack can deprive 3D translation and rotation data of the bones, but the object is limited to single joint and is unnatural in traveling direction. Our method can enlarge its object to multiple joints. Getting more data, we hope we can analyze bone motions including rotations and resultant force etc. more accurately and more practically