신경망을 이용한 지능형 상∙하지 재활 휠체어 로봇 시스템 설계 및 구현

Abstract

ABSTRACT As a way to help the old and the disabled lead an independent life and to enhance the quality of their life, there are active studies conducted on rehabilitation robot that helps to rehabilitate defective functions. For this reason, in this thesis, the characteristics and the mobility of the old and the disabled were taken into consideration. This thesis seeks to propose an intelligent wheelchair robot system with a new concept that can help the old and the disabled to take an exercise and receive rehabilitation training in daily life to properly move the upper and lower libs extremities and to stand up. This will prevent the degeneration of body functions in the minimum and of musculoskeletal. The purpose of this suggested system is to first allow the old and the disabled to perform all movements in daily life independently as much as possible and to minimize the degree of secondary functional impairment. In other words, this research aims to establish a system, which makes it possible for the old and the disabled to perform rehabilitation exercises. Subsequently, it will use the rehabilitation robot system based on the wheelchairs for the upper and lower limbs extremities, in order to diagnose them in a quantitative way on a regular basis and to verify its effectiveness. The basic structure of the suggested system consists of electric wheelchairs for movement, a robot system of the rehabilitation of the upper and the low limbs extremities, and a nerve-network, which is a robot-controlling system for rehabilitation exercises customized for a user based on the user's muscular strength and bio-signals (pulse, breath). The embodiment of the rehabilitation training function of the upper and the lower limbs extremities consists of the upper and the lower libs extremities rehabilitation robot system and the nerve-network controller. The upper and lower limbs extremities rehabilitation robot system is made of rehabilitation 2-DOF and the lower limbs extremities rehabilitation 1-DOF, which were realized as all-in-one wheelchairs. The establishment of the nerve-network controller made it possible to control the system through intelligent talk depending on the user's muscular strength and bio-signals. For the intelligent rehabilitation exercises of the upper and the lower limbs extremities of the suggested system, a user's muscular strength and bio-signals (pulse, breath) were measured and learned through the nerve network, and the talk output was controlled, according to the user's muscular strength and bio-signals, and it was created for customized rehabilitation exercises of the upper and the lower limbs extremities for the purpose to improve the user's rehabilitation approach. Furthermore, the user-friendly interface was implemented so that the data of body (pulse, breath, electromyogram) measured by a tablet PC could be analyzed in real time, which made it possible for a user to monitor the changes in bio-signals following rehabilitation. It can interwork with a smart phone, which improves the efficiency of rehabilitation exercises. The usability of the system proposed in this thesis was enhanced through a drive and a rehabilitation mode. Also, a follow-up performance was analyzed for the evaluation of the performance of the offered algorithm and the verification of its system. In the experiment to assess the performance of the proposed algorithm, the subjects' bio data were first obtained, learned, and analyzed by the nerve-network algorithm. In addition, in order to minimize the influences of the learning data on the experiment results and to secure reliability, 10-Fold cross-validation analysis was carried out. During the experiment, when the subjects took a rehabilitation exercise, the average accuracy of the classification of the talk output depending on the users' muscular strength and bio-signals turned out to be 87.8%. The experiment results showed that the average error in the angle of joints of the upper limbs extremities rehabilitation exercise was 2.52 degrees. The average error in the angle of joints of the lower limbs extremities rehabilitation exercise was 2.46 degrees, and of the standing exercise was 2.14 degrees. The follow-up performance was good enough to take rehabilitation exercise. Therefore, the system proposed in this thesis can guarantee the mobility of its users. The users can continuously take rehabilitation exercise in their daily life, and it offers customized rehabilitation with the data about its users' muscular strength and bio-signals. In conclusion, the system is expected to be helpful to its users, who have the will to rehabilitate and improve the quality of their life by monitoring the data about body and rehabilitation.