草原公路驾驶员脑电信号的降噪处理研究

发布时间：2019-03-04 09:45

【摘要】：内蒙古自治区草原公路路侧景观环境单调,道路线形长直线多曲线半径大,驾驶员在草原公路上行车很容易产生疲劳和注意力下降等情况,对整个道路交通安全造成严重隐患。脑电波作为评定中枢神经系统变化的指标之一,可以比较灵敏得反映驾驶员在行车过程中的精神状态。目前,驾驶员脑电信号的采集方式为多以静态采集方式为参考,由于颅骨和头皮组织等对脑电信号的衰减作用,加之驾驶环境极其复杂并呈现动态变化特征,脑电信号很容易被无关伪迹干扰。因此在分析脑电信号时,必须剔除来自脑电活动以外的伪迹,对获得逼近驾驶员驾驶真实状态的脑电信号很有必要。本文从驾驶员脑电信号伪迹的来源着手,对伪迹特征进行逐个分析并进行综合分类,确定脑电信号中的伪迹主要由脑电信号频带外的高频噪声和频带内的眼电伪迹组成,通过行车试验设计采集到驾驶员的脑电信号数据。对于高频噪声的去除本文设计了 Kaiser窗滤波器和等波纹优化滤波器,通过对指标比较判断,选择最优滤波器。对于频带内部的眼电伪迹,在选择合适的小波基之后,通过对比固定阈值法和改进的各级独立阈值分段阈值法,比较信号在降噪之后的时域和频域图谱分析降噪前后的变化情况。本文研究得到以下结论:(1)通过对比两种方法设计的滤波器对脑电降噪之后的时频图得出:两种方法在滤波之后对脑电信号造成一定程度的延迟效应,但Kaiser窗滤波器对信号在时域造成的时间延迟更长,而且在对高频段信号的滤波达到规定的要求时,Kaiser窗滤波器具有更大的阶数N,造成整个滤波过程计算量变大,增加了数据处理的时间。(2)Kaiser窗滤波器在阻带波纹的不唯一性导致旁瓣泄露明显,等波纹法却在阻带内将第一旁瓣的波动转移到高频部位去,使整个阻带的波动更加均匀,通带内等波纹法使得滤波后的信号与原信号具有更优的逼近程度。(3)通过对比传统的固定阈值法和改进的各级独立阈值分段法处理脑电信号,固定阈值法降噪之后脑电信号丧失了原来信号的细节特征,使得信号过于平滑,并且在眼电伪迹出现奇异点处,波峰存在显著,降噪效果不佳。改进的独立阈值分析能在每级分解中采用多段阈值,降噪后的信号能更好的保留了脑电信号的细节特征,在眼电伪迹处能更好的消除眼电伪迹波形的特征,从频域分析也验证了眼电伪迹的频率被消除。
[Abstract]:The landscape environment on the road side of the grassland highway in Inner Mongolia Autonomous region is monotonous, the long linear and multi-curve radius of the road is large, the driver is easy to lead to fatigue and decrease of attention on the grassland road, which causes serious hidden trouble to the traffic safety of the whole road. As one of the indexes to evaluate the changes of the central nervous system, brain waves can reflect the driver's mental state in the course of driving more sensitively. At present, the acquisition mode of driver's EEG signal is mostly based on static acquisition mode. Due to the attenuation effect of skull and scalp tissue on EEG signal, plus the driving environment is extremely complex and presents dynamic change characteristics. EEG signals are easily interfered with by unrelated artifacts. Therefore, when analyzing EEG signals, it is necessary to remove the artifacts from the EEG activities, which is necessary to obtain EEG signals approaching the driver's real driving state. Starting from the source of EEG artifacts in drivers, this paper analyzes and classifies the artifact features one by one, and determines that the artifacts in EEG signals are mainly composed of high-frequency noise outside the band of EEG signals and eye artifacts in the frequency band, and that the artifacts in EEG signals are mainly composed of high-frequency noise outside the band and eye artifacts in the frequency band. The data of electroencephalogram (EEG) of drivers were collected by driving test design. In this paper, the Kaiser window filter and the equiripple optimization filter are designed to remove the high frequency noise. The optimal filter is selected by comparing the indexes. After choosing the appropriate wavelet basis, the fixed threshold method and the improved independent threshold segmentation method are compared to each other after selecting the proper wavelet basis for the eye electrical artifact in the frequency band. The time domain and frequency domain spectra of the signal after denoising are compared to analyze the changes of the signal before and after de-noising. The conclusions of this paper are as follows: (1) by comparing the time-frequency images of EEG de-noising with the filters designed by the two methods, the results show that the two methods have a certain degree of delay effect on EEG signal after filtering. However, the time delay caused by the Kaiser window filter in the time domain is longer, and when the filtering of the high frequency signal reaches the required requirements, the Kaiser window filter has a larger order N, resulting in a larger amount of computation in the whole filtering process. The time of data processing is increased. (2) the non-uniqueness of the Kaiser window filter in the blocking band leads to obvious side lobe leakage, but the first side lobe fluctuation is transferred to the high frequency region in the blocking band by the equal ripple method, which makes the fluctuation of the whole stop band more uniform. The passband isoripple method makes the filtered signal have a better approximation with the original signal. (3) the EEG signal is processed by comparing the traditional fixed threshold method and the improved independent threshold segmentation method. After noise reduction by the fixed threshold method, the EEG signal lost the detail characteristics of the original signal, which made the signal too smooth, and at the singular point of the false trace of the eye, the wave peak was significant, and the noise reduction effect was not good. The improved independent threshold analysis can adopt multi-segment threshold in each decomposition, and the de-noised signal can retain the details of EEG signal better, and can eliminate the characteristic of eye artifact waveform better at eye artifact. Frequency domain analysis also verifies that the frequency of eye artifacts is eliminated.
【学位授予单位】：内蒙古农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.25

【参考文献】