| Summary: | 博士 === 國立成功大學 === 機械工程學系碩博士班 === 93 === Recently, the development of the prosthesis system has improved the daily life of amputees or patients with motor function diseases. For the claims of more functional and powerful prostheses, some innovative designs are proposed and emphasized the use of biomedical signal. The advantage of using the biomedical signal, especially for the prosthesis application, is that person can communicate with the machine system by the biomedical signal served as the man-machine interface. For example, amputees may control their prosthesis by controlling their residual muscle exertion. Therefore, the purpose of this study is to investigate the feasibility of using two kinds of biomedical signals, electromyogram (EMG) and electroneurogram (ENG), to be the control source of prosthesis. This study focuses on the methodology as to process the two biomedical signals, to extract the kinematical information, e.g. voluntary joint torque or joint angle during movement, from the signals and carry out the real-time control for prosthesis application.
In this dissertation, an innovative control algorithm for prostheses has been developed by using the EMG signal. The triceps and biceps EMG signals are measured and used to estimate the elbow torque. The target of the EMG control algorithm is to control the external assistive torque proportional to the estimated elbow torque. To demonstrate the feasibility of the control algorithm, this study develops an assistive torque system which uses homogenic EMG signals to improve the elbow torque capability of stroke patients by applying an external assistive torque. To simplify the control algorithm, the ratios of the unilateral EMG signals to the elbow torque under isometric contraction at various elbow angles and torque levels are calculated and arranged as a mapping matrix. The applied assistive torque is proportional to the difference between the weighted EMG signals of the biceps and triceps determined by interpolating the mapping matrix. The overall stability of the assistive system is enhanced by the incorporation of a nonlinear damping element within the control algorithm which mimics the physiological damping of the elbow joint and the co-contraction between the biceps and triceps. Adaptive filtering of the control signal is also employed to achieve a balance between the bandwidth and the system adaptability so as to ensure a smooth assistive torque output.
The results of tracking experiment demonstrate the ability of the assistive torque system to assist all the able-bodied subjects and stroke subjects in performing a number of tracking movements with reduced effort of agonist and with no sacrifice in elbow movement performance.
About the study of ENG-based control algorithm, this study proposes an innovative design of neural prosthesis to implement autologous afferent sensing and electrical stimulation on nerves for ankle position control. The ENG signal has smaller amplitude, i.e. 10 uV, than the EMG signal. Technical difficulties involved in applying this approach to motor function restoration require developing techniques to extract useful, stable and repeatable signals and eliminating the artifacts induced by the electrical stimulation and nearby muscle activation. In this study, a multi-channel cuff electrode has developed and implanted around the peripheral nerve to measure the ENG signal. A computer-controlled pneumatic-driven dynamometer is designed to perform passive stretching of a rabbit’s ankle in order to minimize electrical disturbance from the control system under small ENG conditions.
Three technologies for implementing the real-time control of neural prostheses have been developed, namely ENG signal separation, joint angle tracing and ENG signal processing on stimulated nerve. First, the ENG signals recorded with a multi-electrode cuff on the sciatic nerve are employed to investigate the possibility of extracting components ascending from the peroneal and tibial nerves. The results show that the signal separation model, which used only two channels of the sciatic recordings, is sufficient to separate the distal afferent components. Second, a simple empirical model is built based on the results of ENG measurement to estimate the peroneal and tibial nerve signals from the angular trajectory of ankle joint. After determining the parameters of the empirical model, an algorithm is proposed to estimated ankle joint angle trajectory by using the ENG signals of the peroneal and tibial nerves. Finally, an ENG signal processing is proposed to extract cleaner ENG signal measured from a stimulated nerve. These technologies provide a basis for the future investigation of biomimetic sensory feedback for functional neuromuscular stimulation (FNS) control of paralyzed limbs. On the other hand, an innovative control mode of electrical stimulation is proposed in this study. The hybrid amplitude/pulse-width modulation (AWM) of electrical stimulation performs better performance in ankle torque control than the traditional amplitude modulation. By the integration above-mentioned signal processing methods and the AWM control mode, a closed-loop control of FNS is realized in this study.
|
|---|
