Development Of Multi-degree Of Freedom Emg Prosthetic Hand Combined With Eeg

At present, multi-degree-of-freedom SEMG artificial hand has become one of the hot topics in the field of rehabilitation medicine and engineering, which has a very important significance on improving the living quality of amputees and promoting the development of rehabilitation services for disabled people.EMG signal (EMG) originated in biological electrical activity of motor units of the neuromuscular system, reflects functional status of the nerves and muscles, and implies a wealth of information on body movement patterns.So,it has been widely used in lots of areas such as muscle jury diagnosis and biomedicine and so on.Electroencephalogram (EEG) is the total reflection of brain cells electrophysiological activities on cerebral cortex and scalp, which is in connection with a great deal of physiological information. The information of controlling limps is acquired by using the methods of EEG signal processing and analysis.Using EEG to control SEMG artificial hand is immature, but it is an attempt that EMG conbined EEG, which is used for the two signal sources of controlling multi-degree-of-freedom SEMG artificial hand.Some research about pickup and de-noise, feature extraction, multi-degree-of-freedom SEMG artificial hand, information fusions and pattern classification of EMG and EEG were done. Many theory exploration and practice were done.The following work and innovations of this paper were made as follow:(1) From the generation mechanism of EEG, this paper illustrates the classification of EEG, and the characteristics of EEG for controlling SEMG artificial hand, besides, studies the lead methods of electrodes and acquisition of EEG and the circuits of de-noise of EEG.And these circuits were designed aiming at the de-noise and drift restrain circuit of EEG and some differences between now and before were showed.(2) The composition of multi-degree-of-freedom SEMG artificial hand were introduced in this paper, which cotained the acquisition and feature extraction of SEMG, the algorithms of pattern classification and the controlling system of the artificial hand.then, the features of the multi-degree-of-freedom SEMG artificial hand with real-time was analysised.Finally,the paper haved pointed out that the use of EMG combinated with EEG is the way to further enhance recognition accuracy of multi-degree-of-freedom SEMG artificial hand.(3) Study on feature extraction based on wavelet packet transform is presented, and algorithm of the signal characteristic of coefficients and sub-band energy based on wavelet packet decomposition were given. The concrete methods include the feature extractions based on wavelet packet transform coefficients and their subband energies. The method above fully takes account of the unstationarity of SEMG and EEG and validity of each extracted feature. Moreover, feature information of between different motions has more obvious differences.(4) Studied on composition the EEG and EMG feature vector and the implementation process of information fusion methods and pattern classification based on back-propagation (BP) neural network were given. With the analysis of the strengths and weaknesses of block selection algorithm and decomposition algorithm of support vector machine, the sequential minimal optimization algorithm (SMO algorithm) was used for information fusion and pattern classification of SEMG and EEG. This method is the special case of decomposition algorithm, and only is to solve an optimization problem with two variables when the the course of each iteration only adjust the corresponding points on two samples of Lagrange multipliers. Its advantage is that the computational complexity has been reduced each iteration of the process, and the optimal solution can get. The experimental results show that the recognition rates of hand's motions are enhanced efficiently over 90% using the information fusion methods based on EMG contained with EEG, which get higher recognition rates and more stability than that in recognizing EMG as the unique signal source traditionally.