Mechanical Design Of Multi-Dof Prosthetic Hand And EMG Signal Control

Because electromyographic (EMG) signals are the ideal control signal of multi-DOF anthropomorphic prosthesis, the research on prosthetic hand has become one of the most active research areas in medical engineering and robotics. A novel EMG signal controlled prosthetic hand is developed. The motions of thumb, index finger and middle finger are identified via EMG signal recognition. Additionally, by analyzing EMG signals of different electrodes, the EMG signals'pattern recognition method is discussed in this paper.Based on the traditional type prosthetic hand, learned from the novel prosthetic hand mechanism and imitated from human hand, a novel prosthetic hand is designed. By kinematic simulation, the fingers'motion and the grasping abilities are testified.Parametric Autoregressive (AR) model is the traditional time-domain EMG signal analyzing method. In this dissertation, two-channel EMG signals are processed under AR model. The pattern parameters are attained and feather vector is established by iteration of least square method. To ascertain the order of AR model, several rule functions are used. Eventually, 4 order AR parameter model is considered to be the best choice. Collected by Otto Bock electrodes, the experiment results reveal that AR model is able to achieve acceptable EMG pattern identification effects.A time-frequency-domain EMG signal analyzing method which combines Short Time Fourier Transform (STFT) and Singular Value Decomposition (SVD) is purposed in this dissertation. Collected by B&L electrodes, the STFT spectrum of two-channel EMG signals proves that the motions of different fingers have great differentiations. By Hamming Function, the SVD of STFT spectrum matrix obtains distinguishable pattern vector of EMG signals. The comparison of AR model and STFT shows that STFT recognize complicated EMG signal better. AR model is simpler and fits for calculation in DSP.