الفهرس 
IntroductionOnly 14 pages are availabe for public view
Abstract
Surface Electromyography (sEMG) signals have a lot of biomedical applications and modern human-machine interactions. sEMG signals received from muscles that require advanced methods for detection, pre-processing, and classification. Current research technologies are focused, principally on deep neural network architectures that collect spatial data from sEMG signals. Low-cost traditional prosthetic leg, available worldwide, can make walking and stair climbing possible but still difficult. This thesis presents the hardware implementation to the sEMG Powered Prosthesis Actuation (PPA) system using recurrent neural network (RNN) model based on three models long-term short-term memory (LSTM), Convolution Peephole LSTM and gated recurrent unit (GRU), which are used to train sEMG benchmark databases, and find the correlation between the input (sEMG) and outputs (gesture). The following techniques were evaluated by calculating the success of a variety of variables like training time, accuracy loss and hyper-parameters which were applied on eight benchmark datasets, in order to demonstrate the validity of these models, with prediction accuracy at almost 99.6 %. The data were collected from benchmark datasets describing different subjects during performance, and analyzing various gait patterns were used to construct the neural network and to alleviate significant model errors in a real-time setting. Processing circuits, interfacing the output with the controller board, signal amplification, motor driving circuit and single-board computer programming are included in the implementation.
Keywords — sEMG, Recurrent Neural Network, LTSM, Pattern Recognition, RNN, Long-Short Term Memory, Prosthesis leg