基于盲源分离的脑电信号分析技术研究

发布时间：2018-07-17 04:55

【摘要】：脑电信号中包含丰富的信息，经常应用在工程的脑机接口以及临床的疾病诊断中。能否很好地对脑电信号进行分析、准确快速地提取信息决定脑电信号在应用过程中性能的好坏。本文在对国内外的脑电信号分析方法进行研究后，利用近年来信号处理领域中应用较广泛的盲源分离算法，分别以P300脑电信号和运动想象脑电信号为对象，对脑电信号进行分析。 P300脑电信号强度很弱，容易受到环境以及眼动伪迹、心电、肌电和自发脑电信号的干扰，淹没在采集脑电信号中。为了快速高效地将P300脑电信号与各种干扰分离开来，本文通过分析P300时域、频域和头皮空间域的特点，提出以相干平均、小波变换与盲源分离相结合的算法，从时频域和头皮空间域对P300脑电信号进行提取。并且提出一种可以自动地从盲源分离获得的多个源信号估计分量中选取P300对应分量的方法。设计实验对比分析三种盲源分离算法——Informax、FastICA和AMUSE在P300脑电信号提取过程中的性能。通过实验表明，基于盲源分离的P300脑电信号提取方法相比于仅从时频域进行提取的性能有显著提高。针对如何准确有效地提取运动想象脑电信号特征的问题，本文通过分析运动想象脑电信号时域、频域和头皮空间域的特征，提出以小波变换为预处理，并利用二阶盲辨识(SOBI)算法和信息论特征提取(ITFE)算法相结合获得的空间滤波器，从时域、频域和头皮空间域提取运动想象脑电信号的方法，并将能量作为特征。通过实验表明基于盲源分离的运动想象脑电信号特征提取方法具有一定优越性，，盲源分离算法SOBI与ITFE相结合获得的空间滤波器能够反映更真实的大脑源活动。
[Abstract]:EEG signals contain a wealth of information and are often used in engineering brain-computer interfaces and clinical disease diagnosis. Whether the EEG signal can be well analyzed and the information extracted accurately and quickly determines the performance of EEG signal in the process of application. After studying the methods of EEG analysis at home and abroad, this paper uses the blind source separation algorithm, which is widely used in the field of signal processing in recent years, to take the P300 EEG signal and the motion imaginary EEG signal as the objects, respectively. The intensity of P300 EEG signal is very weak, which is easily disturbed by environment, eye movement artifacts, ECG, EMG and spontaneous EEG signals, and is submerged in the collection of EEG signals. In order to separate P300 EEG signal from all kinds of interference quickly and efficiently, by analyzing the characteristics of P300 time domain, frequency domain and scalp space domain, this paper proposes an algorithm combining coherent averaging, wavelet transform and blind source separation. The P300 EEG signals were extracted from time-frequency domain and scalp spatial domain. A method for automatically selecting P300 corresponding components from the estimated components of multiple sources obtained by blind source separation is proposed. The performance of three blind source separation algorithms, Informax-FastICA and AMUSE, in the process of P300 EEG signal extraction is compared and analyzed. Experiments show that the performance of P300 EEG signal extraction method based on blind source separation is significantly improved compared with that of extracting P300 EEG signal from time and frequency domain only. In order to solve the problem of how to extract the features of motion imagination EEG accurately and effectively, this paper proposes wavelet transform as the preprocessing method by analyzing the features of motion imaginary EEG in time domain, frequency domain and scalp space domain. The second order blind identification (SOBI) algorithm and the information theory feature extraction (ITFE) algorithm are used to extract the motion imaginary EEG signals from the time domain, frequency domain and scalp spatial domain, with energy as the feature. Experiments show that the feature extraction method based on blind source separation has some advantages, and the spatial filter based on SOBI and ITFE can reflect more real brain source activity.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.6

【参考文献】