Humanoid Robotic Hand (HRH) Based on EMG signal for Amputees Persons

This paper proposes a system that designed and implemented to be used for persons who lost their limbs because of accident, wars or any other diseases. The idea of this paper is to design and implement Humanoid Robotic Hand (HRH). The HRH was built by using 3D printer technology of hard Polylactide (PLA) filament. The different components of the hands were separately built and then assembled, which gives easy manufacturing at the same time great latitude to choose materials, and also built by six servo motors while five servo motors used to move the fingers and the sixth servo motor used to rotate the wrist. The designed HRH system was controlled by electromyography (EMG) signals were utilized to classify seven classes of movements in offline mode and five movements in real-time. The EMG signals were measured by using three surface electromyography (sEMG) MyoWare muscle sensors, which were located on the forearm on three muscles (Extensor Carpi Ulna, Extensor Carpi Radius and Extensor Carpi Digitorum) and use Arduino Mega microcontroller as an analogue to digital converter to take the signal from the sensor and also use data collector to control the humanoid robotic hand. The proposed pattern recognition system was investigated in an offline mode to enhance it and to develop the classification accuracy of the system by using the (Integral Absolute Value (IAV), Mean Absolute Value (MAV) , Root Mean Square (RMS), Waveform length (WL), Zero Crossing (ZC), Slope Sign Change (SSC) and Autoregressive (AR) as feature extraction , Principal Component Analysis (PCA) as feature reduction, k-Nearest Neighbor (k-NN) and Linear Discriminant Analysis (LDA) algorithms as classifiers. Furthermore, the effects of electrodes' position on the forearm and the number of channels on the efficiency of the pattern recognition system were investigated too. The results showed that the performance of the LDA is better than the k-NN because the accuracy of LDA is 91.1056% and the accuracy of k-NN is 87.5849% these percentages are in the offline mode and in real time mode the accuracy is 84% when using LDA algorithm