Research In EEG Signal Analysis Using LMD Based On Wavelet Packet

EEG signal contains a large number of all the activities of the human spirit, however, eeg is a kind of signal which is nonlinear, nonstationary, and randomnes s is very strong, brought huge difficult to eeg feature extraction. Existing eeg acquisition is generally used scalp eeg acquisitio n method, the collected signal are easily disturbed by the power frequency signal from acquisition machine and electromyograp hic signal and eye electrical signals from human beings themselves, Affect EEG signal feature extraction effect. So, before for eeg feature extraction, the first thing is to remove noise from EEG signal. Wavelet packet denoising method has higher resolution in high frequency and low frequency part, it is a powerful tool for EEG signal denoising which contains a large number of mutations. After pretreatment on EEG signal denoising, researchers have proposed the time domain, frequency domain analys is and time-frequ enc y analysis of EEG signal eigenvalue extraction methods. The time-frequ enc y analysis method can be combined with time domain and frequency domain informatio n, multi-d imens ional analysis of EEG signals, has been widely applied. In the time-frequency analysis method, the local mean decomposition is a effic ien t eeg signals time-frequency analysis method,can decompose EEG signal into a series of product function component, and in the process of calculating product function component, can very convenient to calculate the weight of amplitud e and frequency informatio n, greatly reduce and simplifies the workload of eeg feature extraction, the algorithm is more quick and convenient.Our acquisitio n of eeg signals contain more noise, the existing eeg feature extraction algorithms take too much time when the eeg signals are decomposed, they are not suitable for building real-time systems. This paper proposes a loc al mean decomposition method based on wavelet packet. Try to before for EEG feature extraction, for effective denoising of EEG signals, and using the loc al mean decomposition approach to EEG signal decomposition. In this paper, main work is as follows:Firstly, we studied the mechanism of eeg signal nonlinear, nonstationary and randomness is strong, comparative study the time domain, frequency domain analysis and time-freq uenc y analysis methods and so on existing eeg analys is method.Secondly, we studied the common brain electrical signal denoising method, wavelet denoising and wavelet packet denoising, analysis of two kinds of denoising method principle, is proposed based on signal to noise ratio, root mean square error and the correlation coefficient denoising evaluatio n index, thro ugh the experiment proves the advantage of wavelet packet denoising. The wavele t packet denoising can lead to further analysis of the high frequency part of signal, wavelet packet denoising has better denoising effect to EEG signals of which has a lot of mutations.And then, we discusses the existing time-frequency analysis method of empirical mode decomposition and the ensemble empirical mode decompositio n method, the modal aliasing and the boundary effect and the problem of lo w algorithm efficienc y of decomposition and its mechanism. The empirical mode decomposition has modal aliasing problems and boundary effect,Due to its mean value function is through the maximum and minimum structure fluctuatio n envelope respectively, Low density in border efficient point, lead to serious boundary effect. In addition, the ensemble empirical mode decomposition metho d using multip le join white noise empirical mode decomposition method for many times, although to a certain extent solved the empirical mode decomposition of modal aliasing problems, however, empirical mode decomposition result in its many times for the same period of the signal decomposition time exponentia lly, not suitable for real time analysis system of the building.Finally, using the local mean decomposition method based on wavelet packet to analyze the EEG signals. Local mean value a-nalysis method, the use of adjacent extreme value point of the mean average structure function, on the border of efficient point density is twice that of empirical mode decomposition method, effectively alleviate the problem of boundary effect, at the same time, the decomposition efficiency is higher. Through experimental comparison and analysis, local mean decomposition based on wavelet packet method compared to the empirical mode decomposition and the ensemble empirical mode decomposition method, can effectively alleviate the boundary of the empiric a l mode decomposition effect, at the same time decomposition efficiency is muc h higher ensemble empirical mode decomposition method, and decompositio n efficiency is basically the same with empirical mode decomposition method.