Musculoskeletal Signal Acquisition And Its Application In The Control Of Prosthetics

Compared with other biological signal, Mechanomyography(MMG) signal is a sound by body’s muscle contractions. MMG signal not only has many advantages such as strong adaptive capacity to environment, good interference immunity, good anti-jamming ability, good anti-fatigue ability and low cost, but also contains abundant muscle activity information. So it can be used as prosthetic control signal source to achieve the multi-degree control of intelligent prosthetic hand, especially in rehabilitation engineering about intelligent prostheses. A number of domestic and international researches show that MMG signal has an excellent prospect in intelligent prosthetic hand control and expresses a great application prospect, so intelligent prosthetic based on MMG signal has become a new research topic. This paper’s subject is “MMG signal acquistion and its application research on prosthetic hand control”. The MMG signal is used as the prosthetic control signal source in the paper, which is collected at the specialized forearm-muscle by MEMS three-axis sensor, and is classified to recognize the different hand gestures with algorithm. Finally, the developed prosthetic system is used for validating.The main contents and innovations in this paper are as follows:(1)、The research subject selects MEMS three-axis sensor to collect the MMG signal and it is produced by Xsens. The MMG signals are typical nonlinear and non-stationary. The some interference signals, overlapping the effective MMG signal in frequency spectrum, are easy to mix into acceleration signal and are difficiulty to be removed. In deep study of the limatations of the traditional digital filter, the paper focuses on the empirical mode decomposition(EMD) as a new signal processing method and filters the noise combining with the the Chebyshev digital filter. It makes a algorithm verification about the character which the EMD is suitable for nonlinear and non-stationary signal.(2)、As the traditional neural networks exist shortcomings which is slow and easy to fall into local minimum value in hand gesture recognition., this paper designs an isolated MMG signal hand gesture recognition model which based on a deep belief network constructed by Restrictions Boltzmann Machine(RBM). The model first trains the everyone RBM individually until the last RBM. Then, all trained RBMs are stacked into the deep belief network, and the deep belief network is optimized by the back-propagation algorithm, so it is a trained optimal deep belief network model. Finally, the wavelet coefficient energy of the every channel signal as a input of the deep belief network which is an isolated hand gestures recognition. After making simulation experiments, this system achieves a higher recognition rate comparing with the improved BP neural network, the time efficiency of makes a good improving and the model indeed prevent over-fitting.(3)、Aiming at verifying the validation of experimental result and algorithm model, the paper develop the experimental platform. The real-time software developed by QT5 platform can collect the MMG data and classify the MMG signal to get the hand gesture feature information. The PC automatically transfers the feature information to the prosthetic hand platform by USB-422 serial port. Prosthetic hand platform decode the signal and drive the prosthetic hand to perform the appropriate action. Eexperiment show that the prosthetic hand system platform can classify the hand gesture and the research’s practicability is confirmed.