No abstract

No abstract

No abstract

In this paper, both hardware and software design to develop a wearable walking monitoring system for gait analysis are presented. For hardware, the mechanism proposed is adaptive to different individuals to wear, and the portability of the design makes it easy to perform outdoor experiments. Four force sensors and two angle displacement sensors were used to measure plantar force distribution and the angles of hip and knee joints. For software design, a novel algorithm was developed to detect different gait phases and the four gait periods during the stance phase. Furthermore, the center of ground contact force was calculated based on the relationships of the force sensors. The results were compared with the VICON motion capture system and a force plate for validation. Experiments showed the behavior of the joint angles are similar to VICON system, and the average error in foot strike time is less than 90 ms.

The EMG signal is a well-known and useful biomedical signal. Much information related to muscles and human motions is included in EMG signals. Many approaches have proposed various methods that tried to recognize human motion via EMG signals. However, one of the critical problems of motion pattern recognition is that the performance of recognition is easily affected by the normalization procedure and may not work well on different days. In this paper, a modified feature of the multi-channel EMG signal is proposed and the normalization procedure is also simplified by using this modified feature. To recognize motion pattern, we applied the support vector machine (SVM) to build the motion pattern recognition model. In training and validation procedures, we used the 2-DoF exoskeleton robot arm system to do the designed pose, and the multi-channel EMG signals were obtained while the user resisted the robot. Experiment results indicate that the performance of applying the proposed feature (94.9%) is better than that of conventional features. Moreover, the performances of the recognition model, which applies the modified feature to recognize the motions on different days, are more stable than other conventional features.