【作者】 彭宏；

【导师】 胡斌；

【作者基本信息】 兰州大学 ， 无线电物理， 2015， 博士

【摘要】 随着新型传感材料与低功耗微型处理器研究的发展,生物传感器的应用越来越普遍,已逐渐被应用于军人、宇航员、运动员、病人等生理信息监测中,因此,基于可穿戴生物传感器的生物信息感知理论与技术研究的重要性日益突出。而大脑作为人体的最重要控制中枢,对脑电信息的感知成为生物信息感知研究中的重点。由于精神疾病、认知能力等与大脑之间的紧密关联性,使得基于可穿戴脑电传感器的脑信息感知研究成为当前的热点之一。本文以可穿戴脑电信息感知研究为核心,面向脑电信息感知与处理中的几个关键问题,研究了相应的解决算法和模型,并以欧盟框架7国际合作项目“心理疾病在线预测与干预系统研究”为依托,验证了所提出的算法和模型在与精神压力状态相关的生物信息感知系统中的有效性。由于脑电信号极其微弱,在科研单位或医院都会设立特定的屏蔽室来排除干扰,并由专业人员操作脑电采集仪以保证信号的可靠。而在普适化环境中,用户需自主操作佩戴脑电传感器,其周围存在不可避免的干扰,因而需要对脑电信息感知中的新问题做进一步研究。本文研究了普适化脑电信息感知中的三个关键问题,创新性研究成果包括如下三个部分：(1)一种新型的脑电信号质量评估模型：当用户脱离了医院的环境而自主操作脑电传感器时,如果传感器电极没有正确连接在额头上或所处的实验环境电磁噪声较大,则所采集到的信号不可使用,会导致实验的失败。因此,需要研究快速有效地评估脑电信号质量的算法模型来指导使用者判断数据可用性。而当前对脑电信号质量评估的算法非常匮乏,大都集中在简单的皮肤-电极阻抗上面。本文基于模糊综合评价法的思想,通过大量的数据分析,结合专家意见评判,提出了一种新的模型来评估原始脑电图信号质量。该模型运行速度极快,可脱离计算机环境而工作于脑电传感器中,因而方便用户的自测。在实际项目执行中验证了本算法模型的有效性,解决了项目初期实验数据质量控制的难题。(2)无参考电极时较少导联下脑电信号中眼动噪声的去除算法：可穿戴传感器的电极安置在前额,距离眼睛非常近,因而所采集到的脑电信号容易受到眼动噪声的污染。本文结合眼电与脑电的差异性特征,研究了两种可用于较少导联环境的眼动噪声去除算法：一种为小波变换结合独立分量分析的算法,利用小波变换构造出眼电参考信号,之后通过独立分量分析算法提取出真实眼电信号；另外一种为小波变换结合自适应滤波算法,由小波变换法重构的眼电信号作为自适应噪声抵消器的参考输入信号,利用自适应滤波器的跟踪特性去除眼电噪声。这两种方法适用于较少导联下便携式脑电采集环境,在滤除眼动噪声的同时能够最大限度地保留脑电信息。在本项目的数据处理中验证了算法的有效性,得到相对纯净的脑电信号,为后期的特征分析打下基础。(3)精神压力状态与脑电特征的关联性：脑电特征的分析与特定应用环境相关,本文依托项目的研究目的在于利用感知到的生物信息评估用户是否承受着较高的精神压力,因此与精神压力状态与脑电特征量的关联性探索成为研究关键。基于本文自主设计的可穿戴脑电传感器,在中国、西班牙和瑞士三个国家中的受试人群中采集实验数据(自评量表与脑电数据),利用最近邻与朴素贝叶斯分类算法,对脑电的特征进行分类研究。在传统的线性特征之外,找到了与精神压力相关的非线性脑电特征。本文的研究意义在于精神疾病的早期发现与干预。由于社会发展节奏的加快,导致心理与精神疾病的发病率持续升高,因而带来巨大的医疗负担和经济、社会损失。而目前各国普遍存在专业医护人员缺乏、医疗方式有限、时效性差等各种不足。“心理疾病在线预测与干预系统研究”基于可穿戴的生物传感器系统感知人体的生物信息,结合电子化问卷和量表等,以计算机化的认知行为疗法为辅助,研究心理或精神疾病发病风险,提供预警和在线干预。基于项目背景,本文研究了普适化脑电信息感知中的关键问题,基于所开发出的可穿戴脑电传感器,结合以上三个方面的研究成果,应用在“心理疾病在线预测与干预系统研究”中,辅助实现了精神压力的诊疗系统,具有一定的应用参考价值。更多还原

【Abstract】 With the development of novel sensor materials and low-power microprocessor research, biosensors have been widely applied to physiological information monitoring on soldiers, astronauts, athletes and patients. Thus, bioinformatics perception theory and technology based on wearable biosensors have shown increasing importance and gained considerable development. Since brain is the most important control center of body, research of EEG(Electroencephalogram) analysis becomes the focus of bioinformatics perception study. Most of mental illnesses and cognitive abilities are closely associated with brain, so the research of EEG based on wearable sensor has become a hot spot. This paper focused on methodology of EEG data sensing and awareness, proposed model and algorithm to solve several key questions in EEG information processing. Based on the Seventh Framework Programmem of EU "Online Predictive Tools for Intervention in Mental Illness," the model and algorithm are proved to be efficient in bioinformatics sensing systems.The EEG signal is so week that many institutes and hospitals set special screened room where professional persons operate the machines. But in the ubiquitous environment filled with noises, users have to do it themselves. It’s important to research the key problems of EEG data sensing and awareness related to ubiquitous environment. This paper solved key questions in the new methodology and proposed 3 innovative parts:(1) A novel assessment model of EEG signal quality:When the user operates EEG sensor out of the hospital environment and the sensor electrode is not properly attached to the forehead, the EEG signal collected then will be not available, resulting in failed experiment. Therefore, it’s needed to design a quick and effective way to assess the quality of the EEG signal so as to guide the user to determine EEG data availability. But the algorithm to assess the quality of the EEG is very scarce right now, and most reaearches concentrated in the skin-electrode impedance. In this paper, a novel assessment model to assess the raw EEG signal quality is proposed which is based on the idea of fuzzy comprehensive evaluation, large amounts of data, and experts’opinion evaluation. This model can work in EEG sensor fastly and thus users can do self-test. Thus the model is proved to be efficient and solves problem of EEG signal quality control in begin of project.(2) EOG noise removal algorithm under less leads EEG:Electrodes of wearable EEG sensor are placed on the forehead, where is close to eyes. Thus the EEG is susceptible to EOG(Electrooculogram). This paper studied two EOG noise removal algorithms with less leads of EEG. One is a combination of independent component analysis and wavelet transform. Wavelet transform is used to constructed EOG reference signal, with which the independent component analysis methord works to extract the real EOG noise. Another algorithm is a combination of Adaptive Noise Cancellation and Discrete Wavelet Transformation, the reference of EOG is reconstructed by Discrete Wavelet Transformation and used as input of Adaptive Noise Cancellation to remove real EOG noise. Both of them are suitable for small portable EEG sensor works under less leads to remove the EOG and keep real EEG, and have been widely used in data processing in this project and proved to be efficient.(3) Analysis of EEG features related to mental stress:EEG features are related to specific application. The purpose of this project is to take advantage of perceived biological information to assess whether someone suffers mental stress, so the exploration of stress-related EEG features is critical. Based on the experimental data (Scale and EEG data) from China, Spain and Switzerland, this paper classifid the characteristics of EEG with KNN and naive Bayes classification algorithm to find the correlation between mental stress and EEG. Besides traditional linear features, some new non-linear features are worked out related to mental stress.The aim of this paper is find out and intervene mental stress in the beginning. In this highly competitive and fast-tempo society, incidence of psychological and mental illnesses continues to rise up, leading to tremendous healthcare burden and economic losses. At the same time, various deficiencies such as the lack of health care professionals, poor timeliness of limited medical treatment methods, are very popular in many countries. The project of "Online Predictive Tools for Intervention in Mental Illness" tries to collect the body’s biological information with wearable biosensor system, get their electronic questionnaires and scales, and with help of computerized cognitive behavioral therapy, to research the risk of psychological or psychiatric illness, and provide early warning and online intervention. Based on the novel wearable EEG sensor combined with the results above, this paper studied the key problems of ubiquitous EEG data sensing and awareness, and put it to project of "Online Predictive Tools for Intervention in Mental Illness" to achieve a stress clinic system to contribute certain social significance.更多还原