Exoskeleton hand

In this thesis, hand exoskeleton system is designed to provide rehabilitation and support finger movements for patients who have lost some or all of their hand functions due to many reasons such as stroke, stroke, spinal cord injuries, tendon injuries. The hand exoskeletal system, which designed using an electric actuator, has a structure that is placed on the front (dorsal) part of the hand and actively supports 5 finger movements. In the design, there are 14 active extension / flexion degrees of freedom, 2 of which are in the thumb and 3 for the other fingers. It also has a total of 6 passive degrees of freedom to create abduction / adduction and extension / flexion movements in the wrist for each finger. In addition, the design has 6 passive degrees of freedom to create abduction / adduction for each finger and extension / flexion movements for wrist. The skeletal system was designed with 3D modeling applications by taking the anthropometric dimensions and degrees of freedom of the hand as reference. To print the design, PLA (polylactic acid) filament is used as a light, durable, flexible and completely biodegradable. The light and portable prototype produced with a 3D printer provides the opportunity to assist the patient in home rehabilitation and daily life activities. Control of the hand exoskeleton is provided with the android application developed for the designed system. The hand exoskeleton system, which performs rehabilitation exercises defined by the recommendations of the physiotherapist, perceives intent using sEMG, GYRO, ACC and ORI signals to support hand movements and can classify 32 different hand movements with artificial intelligence algorithms in real time, has been designed. In the developed system, intent detection was achieved by using the data received with EMG, GYRO, ACC and ORI sensors with the MYO armband placed on the forearm. EMG signals have been normalized with MVC technique and MAV, STD, VAR, RMS, IEMG, ZC and WL feature vectors have been removed and effective features have been selected. In the classification phase, 5 classifiers created using linear support vector machines (SVM), which is one of the machine learning algorithms, were used in parallel.