【作者】 侯亚培；

【导师】 王金甲；

【作者基本信息】 燕山大学 ， 通信与信息系统， 2012， 硕士

【摘要】 脑机接口BCI是一种新的实现人与外部环境进行通讯的人机交互系统，它不依赖于人体固有的神经和肌肉，只需要利用大脑意识就可以实现与外界的信息交流。脑电信号的特征提取、模式分类是脑机接口研究中的热点问题，但是对脑电信号的去噪预处理也是研究中必不可少的部分。本文运用卡尔曼滤波方法对脑电信号进行去噪处理。首先，针对脑电信号中存在工频干扰的情况，研究了使用卡尔曼滤波方法来消除工频干扰。在文中将工频干扰当成正弦信号来处理，通过正弦信号的和差化积等变化关系建立了卡尔曼滤波模型，从而得到了其运算过程中所需的参数。针对正弦信号变频、变相、变幅值的情况进行了实验，估计了正弦叠加信号的每个分频率的值。然后又针对标准的脑电信号和实际脑电信号进行了实验，实验结果表明，此方法很好的去除了工频干扰。其次，针对脑电信号中受到多种干扰如眨眼、心电、肌电等的问题，实现了一种卡尔曼滤波模型和模型参数估计的方法，仿真实验所用的数据是公开的脑机接口竞赛实验数据(BCI Competition III dataset I)。实验结果表明通过本方法去噪预处理后，分类正确率有了很大的提高，且优于小波去噪与谱减法等预处理方法。最后，对扩展卡尔曼滤波方法进行了研究，以正弦信号为例来进行仿真实验。通过非线性计算得到正弦信号与其观测值，对观测值进行扩展卡尔曼滤波。实验结果表明，该方法很好的还原了原信号。更多还原

【Abstract】 A brain computer interface (BCI) is a new man-machine interactive communicationsystem which can realize communication between people and external environment. Itdoes not pass through the human body inherent nerves and muscles, only need to usebrain consciousness to achieve information exchange with the outside world. Featureextraction, pattern classification of EEG is a hot issue in the study of brain-computerinterface, but EEG de-noising preprocessing is an essential part of the research. In thispaper, we use the kalman filtering method for EEG de-noising.First of all, a method of using kalman filter to eliminate power line interference isresearched to solve the problems of power line interference which is included in the EEG.Power line interference in the text will be dealt with as a sinusoidal signal. Kalman filtermodel is established through variation relationship of the sinusoidal signal, and get therequired parameters which is need in the course of its operation process. We deal with theconversion of frequency、phase and amplitude sinusoidal signal on the experiment, andestimate value of each sub-frequency in sine superimposed signal. And then experimentsabout standard EEG and actual EEG were carried out. Experimental results show that thismethod is good to wipe off power line interference.Secondly, for the issues that EEG is affected by a variety of interference such aseyeblink、cardiac electric activity and myoelectric activity etc, we achieved a method ofkalman filter modeling and model parameter estimation. The open brain ComputerInterface Competition experimental data (BCI Competition III dataset I) is used in theexperiment. The experimental results show that after the de-noising pretreatment of thismethod, Classification accuracy has been greatly improved and is better than thepretreatment method of wavelet de-noising and spectral subtraction etc.Finally, the extended kalman filtering method is discussed. We use a sine signal inthe simulation experiment. We get sine signal and its observation through the nonlinearcalculated, and carry through extended kalman filter to the observations. Experimentalresults show that this method is good to restore the original signal.更多还原