Although many wearable sensors have been developed or commercialized for gait analysis, ambulatory force sensors and motion sensors were rarely integrated to implement complete human dynamics analysis. Three dimensional (3D) lower limb kinematic and kinetic analysis based on ambulatory measurements is introduced in this paper. We developed a wireless sensor system composed of a mobile force plate system, 3D motion analysis units and a wireless data logger. 3D motions of body segment and triaxial ground reaction force (GRF) could be simultaneously measured using the system, and the data obtained from sensor units on thighs, shanks and feet could be transferred to a personal computer by wireless local area network (LAN). A stick-chain model was built to visually analyze lower limb postures and joint trajectories, and an inverse dynamics method was adopted to calculate triaxial joint moments. We conducted application experiments on volunteers to visualize and analyze 3D segment orientations and joint moments during successive gait for level ground walking, treadmill walking, and stair climbing. Experiment results showed that 3D segment orientations and joint trajectories could be shown and analyzed using the stick chain model, and that joint loads for different walking conditions could be compared in each gait cycle.

Numerous rehabilitation robots have been developed to deliver therapy for hemiplegic patients with a unilateral-disabled limb or for aged persons with motor-function degenerated limbs. Recently, much attention has been paid to the development of robots that support bilateral arm training in various patterns. It has been proved that the coordination of two limbs can reduce hypertonia and abnormal synergies compared to unilateral limb training. However, traditional robots realized different modes with the robot providing a corresponding force for the impaired limb. This paper proposes a novel bimanual training system. A subject controls master and slave terminal handles with his/her two limbs. The less impaired limb provides a corresponding force for the more impaired limb to achieve different training patterns. No matter in which mode, the two limbs accomplish symmetric movement. An experimental prototype was built. Frequency response test and training tests on 9 healthy subjects in active-resisted and active-assisted modes were performed. Experimental results confirm that the system can correctly respond a subject's commands if velocity was changed within the frequency range of 30 Hz. For each training mode, the average position error, and RMS (root mean square) values of position errors and velocity errors that between the reference values and actual values of a motion tracking task were reduced after practice. This proves the effectiveness of the system for self-assisted rehabilitation training.

To achieve accurate and efficient interaction with humans, robot training is indispensable to make robot cooperate with different host. We are focusing on development of a human motion analysis system to real-time measure body segment orientations. Imitation control was applied on a biped robot based on measurements of the developed wearable sensor system. Experimental study was implemented to verify the synchronous imitation control method proposed for the biped robot, and verification results proved the feasibility of the proposed control method. Through comparing results obtained from direct imitation control method and improved method based on training algorithm considering the personal motion pattern, we found that the imitation control accuracy was markedly improved, and the three-axial average errors of x- y- and z- moving displacements related to leg length were 12%, 8% and 4% respectively.

A wearable sensory system for human lower extremity motion analysis is proposed, and an intelligent computation method for this sensory system is presented. The standard method for human motion analysis is the optical motion analysis using high-speed cameras to record human 3D motion, but this method is only limited in the laboratory research, because it requires expensive devices, large space and time-exhausted calibration experiments. In this study, two low-cost human motion analysis systems are constructed, deferent from the conventional 3D motion analysis system based on high-speed camera. These wearable systems incorporate gyroscopes (ENC-05EB) to measure angular velocities of body segments, and two-axis accelerometers (ADXL202) are used to measure the accelerations for the purpose of leg (foot, shank and thigh) motion analysis in every human motion cycle. The first wearable sensor system is designed for only foot motion analysis and the second system can be used for a leg (foot, shank and thigh) motion analysis. Then based on the two sensor systems, a fuzzy inference system (FIS) is developed for the calculation of the gait phases fed by sensors outputs. A digital filter is also designed to eliminate noises from of the output of the fuzzy inference system, which enhances robustness of the system. Finally, experimental study is conducted to validate the wearable sensor systems using an optical motion analysis system (Hi-DCam)