Study And Analysis Of EEG Signal Based On Motor Imagery

Today, the Brain-Computer Interface(BCI) technology has become a hot spot for researchers. BCI is different from the traditional ways which depend on the human brain normal controls through the output neural network and muscle tissue, it is a new communication channel which can convert the electrical activity to the control signal of devices such as computers, robots and so on. BCI is a new control system that can help the normal brain but incapacitated persons through imagining movement to achieve direct communication with the outside world. So, it has received widespread attention in the fields of biomedical engineering and medical rehabilitation, and the BCI technology based on the imagined movement electroencephalogram(EEG) has the advantages of simple operation, effective and no danger. It obtains many researchers because of its far-reaching research significance and application value. This article is the study and analysis of EEG signal based on motor imagery. The mainly research content includes three aspects, they are the signal preprocessing, feature extraction and pattern classification, expect the purpose of accurate and realtime identification of EEG signals is reached.Firstly, experiment scheme is created to collect the EEG signals data. For the study of the subject, a new experimental brain consciousness tasks is designed to record the EEG signals data according to the experimental requirements of three subjects based on Shanghai NCC Electronic Co.Ltd NCERP 16 channel EEG collection system, and select the signal of EEG data which work obviously to research, then, combined with the characteristics of EEG signals, finite impulse responds(FIR) is applied to band-pass filtering for 0.5-30 Hz, and the method of total average energy transformation is used to select the key research channels, it basically can reduce the interference that the high frequency of muscle electrical, but also eliminates the interference of power frequency components, meanwhile, combined with the anatomy and function of the brain, the select channels are effective and reasonable, and have got good results in reducing the workload and difficulty of next study, and then, for the feature extraction of EEG signals, combined with the chaotic characteristic and the rhythm of sexual energy about EEG, the three kinds of feature extraction method are used to research the characteristics of the test data, they are using chaos analysis to calculate the maximum Lyapunov exponents and correlation dimension, using the discrete wavelet transform to calculate the sum of wavelet sub-band energy and using the wavelet packet transform to calculate the rhythm average energy of wavelet packet and the wavelet packet energy entropy, on this basis two kinds of combination feature extraction method are also given to extract the feature vector, they are chaos analysis combined with wavelet transform and wavelet packet transform, respectively, make full preparation for the next recognition test, finally, for the pattern classification problems of EEG, we use the three kinds classification method of BP neural network, binary tree support vector machine(SVM) and extreme learning machine(ELM) to classify discuss, under the condition of using different feature extraction method to obtain characteristic vector.In the end, from the identification by simulation results shown that, the feature extraction method of chaotic analysis combined with wavelet packet transform, and cooperate with the extreme learning machine classification method have the optimal effect in the solving problem about this subject, and suitable for many kinds of signal analysis.