小波预处理的ITD方法在脑电信号处理中的应用研究

发布时间：2018-01-15 23:18

本文关键词：小波预处理的ITD方法在脑电信号处理中的应用研究　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：脑电信号(Electroencephalogram,简称EEG)属于人体的一种生物电信号,其中包含着反应人类精神状态、情绪活动等的一系列生理信息。对脑电信号的特征进行定性和定量地研究和分析,可以帮助人们进一步认识脑功能并利用脑电信号来做某些疾病的辅助诊断。在癫痫疾病的辅助诊断中,通过对患者发病前、发病中和发病后的脑电信号进行特征提取并做对比分析,可以根据不同的特征频率,更加有针对性地对患者提出诊疗方案。但是,目前在脑电研究领域面临的一大难题是做特征提取时准确性较低,这就使得利用脑电信号做辅助诊断疾病的方法在精确度上存在很大的改善空间。这个问题主要是由脑电信号的两大特征造成的:第一,脑电信号比较微弱且非常敏感,因此在采集过程中极易受到各类噪声的干扰,如采集仪器本身的交流电干扰,被采集者本身的眼电、心电信号干扰等;第二,脑电信号属于非平稳非线性信号,需要采用特殊的非线性信号处理方法来对其进行分解和分析。以往使用的脑电信号分析方法或多或少都存在一些缺点,如小波变换分析法虽然发展较成熟,但是却对脑电信号不具有自适应性;经验模式分解虽然有自适应性,但是却存在较严重的模态混叠和端点效应等问题。这些问题都在一定程度上阻碍着脑电信号特征提取的应用和发展。本论文主要针对上述两方面问题,提出了一种基于小波预处理的固有时间尺度分解(Wavelet-denoising Intrinsic Time-scale Decomposition,简称WD-ITD)方法。该方法首先对含有噪声的脑电信号进行小波去噪预处理,使得处理后的脑电信号信噪比变大,继而使用ITD方法,对脑电信号进行进一步分解。具体研究内容如下:首先,通过对国内外脑电信号研究现状的综述,概述了脑电信号的产生机理、采集方式和特性分类,并对脑电信号的分析方法做了对比和分析。其次,探讨了脑电信号容易受噪声干扰的原因,分析了脑电信号噪声的种类以及常用去噪方法,建立了脑电信号的小波去噪算法评价标准,做了两个对比性实验,验证了小波去噪方法在脑电信号去噪领域的可行性。再次,介绍了一种新的时频分析法——固有时间尺度分解,将该方法与小波去噪相结合形成WD-ITD分解。验证了WD-ITD分解在处理脑电信号方面的可行性和优势,通过与经验模式分解的对比实验,验证了WD-ITD分解在端点效应问题、模态混叠问题以及分解效率方面,都优于经验模式分解。最后,运用WD-ITD方法对癫痫患者的脑电信号做分析。实验结果表明,使用WD-ITD分解后的脑电信号,分解效率平均提高了近4%,分解后得到的固有旋转分量与原始脑电信号的相关度平均提高了3.57%。因此,该方法在有效地解决脑电信号受到噪声干扰问题的同时,能够取得良好的信号分解效果。
[Abstract]:Electroencephalograms (EGG) is a kind of bioelectrical signal of human body, which contains the state of human spirit. A series of physiological information such as emotional activities. The characteristics of EEG signals are studied and analyzed qualitatively and quantitatively. It can help people to further understand the brain function and make use of EEG signal to do the auxiliary diagnosis of some diseases. In the auxiliary diagnosis of epilepsy disease, the patient can be diagnosed before the onset of the disease. The feature extraction and comparison analysis of EEG signals after the onset and the onset of the disease can be more targeted to the patients according to the different characteristic frequency. However. At present, one of the most difficult problems in the field of EEG research is the low accuracy of feature extraction. This makes the use of EEG as an auxiliary method of disease diagnosis there is a lot of room for improvement in accuracy. This problem is mainly caused by the two characteristics of EEG: first. EEG signal is weak and very sensitive, so it is easy to be disturbed by various kinds of noise in the process of acquisition, such as AC interference of the acquisition instrument, Eye-electric and ECG interference of the collector itself. Second, EEG signals are non-stationary and nonlinear signals, which need to be decomposed and analyzed by special nonlinear signal processing methods. For example, although the wavelet transform method is mature, it is not adaptive to EEG signals. Although empirical mode decomposition is adaptive. However, there are some serious problems such as modal aliasing and endpoint effect. To some extent, these problems hinder the application and development of EEG feature extraction. An inherent time scale decomposition based on wavelet preprocessing is proposed. Wavelet-denoising Intrinsic Time-scale Decomposition. WD-ITD method. Firstly, wavelet denoising is used to pre-process the EEG signal with noise, which makes the signal to noise ratio (SNR) of the processed EEG signal to be larger, and then the ITD method is used. The detailed research contents are as follows: first of all, by summarizing the current situation of EEG research at home and abroad, the generation mechanism, acquisition methods and characteristics classification of EEG signals are summarized. And the analysis method of EEG signal is compared and analyzed. Secondly, the reason of EEG signal being easily disturbed by noise is discussed, and the types of EEG signal noise and the commonly used denoising method are analyzed. The evaluation standard of wavelet denoising algorithm of EEG signal is established, and two comparative experiments are done to verify the feasibility of wavelet denoising method in the field of EEG de-noising. A new time-frequency analysis method, the inherent time scale decomposition, is introduced. The method is combined with wavelet denoising to form WD-ITD decomposition, which verifies the feasibility and advantage of WD-ITD decomposition in EEG signal processing. It is verified that WD-ITD decomposition is superior to empirical mode decomposition in endpoint effect problem, modal aliasing problem and decomposition efficiency. The WD-ITD method was used to analyze the EEG signals of epileptic patients. The experimental results showed that the decomposing efficiency of EEG signals decomposed by WD-ITD was increased by nearly 4% on average. The correlation between the inherent rotation component and the original EEG signal is increased by 3.57 on average. Therefore, this method can effectively solve the problem of EEG interference caused by noise. Can achieve good signal decomposition effect.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R741.044;TN911.7

【相似文献】