SIA OpenIR  > 工业控制网络与系统研究室
基于多模脑机接口的神经信息处理和神经反馈方法的研究
其他题名Research on Neural Information Processing and Neurofeedback based on Multi-modal BCI
尹旭贤1,2
导师王志东 ; 石刚
分类号R651
关键词脑电 功能性近红外 脑机接口 神经反馈
索取号R651/Y56/2016
页数127页
学位专业机械电子工程
学位名称博士
2016-11-28
学位授予单位中国科学院沈阳自动化研究所
学位授予地点沈阳
作者部门工业控制网络与系统研究室
摘要脑机接口技术是一种新颖的人机交互方式,该技术不依赖于正常的神经通路,直接利用大脑的想象来控制外部设备,即所谓的“意念控制”。脑电(EEG)是脑机接口中应用最为广泛的一种测量模态,测量的是神经元的放电活动,有较高的时间分辨率,然而空间解析度比较低,因为每个电极测量到的可能是多个脑区的叠加信号。功能性近红外光谱(fNIRS)是一种通过光学测量原理来获得大脑皮层血红蛋白浓度变化的技术,类似于fMRI,空间分辨率较EEG高,但是存在固有的时间延迟,时间分辨率低。两种模态固有的缺陷,也限制了各自在时-空维度上对大脑信息的解析。另外,基于运动想象的脑机接口控制模式较为单一,只能控制机器人的运动方向,并不能在机器人的速度和力上进行控制,这就限制了脑机接口在多维、复杂控制上的应用。最后,用户的操作能力也直接影响到脑机接口的性能,很多情况下,用户不能根据反馈结果及时调整自己的脑活动,从而产生延迟的控制指令,影响控制效果。针对以上出现的问题,我们展开了如下相关研究:1.提高运动想象脑机接口意图识别率的研究。EEG时间分辨率较高,但是空间解析度低,fNIRS空间分辨率高,但存在固有的时间延迟,时间分辨率低。EEG和fNIRS联合测量可以弥补各自的缺陷,在时-空维度上更全面地提取出大脑信息,增加有用的特征量,从而提高运动意图的解码准确率。因此,我们研究了基于EEG和fNIRS联合测量的多模态脑机接口技术,并且自主搭建了多通道fNIRS数据采集平台,研究结果表明,EEG和fNIRS混合模态比单独EEG或fNIRS模式具有较高的意图识别率。2.增加运动想象脑机接口中控制模式的方法研究。传统的基于运动想象的脑机接口常用于机器人等外部设备的方向控制,存在控制指令少、模式单一的问题。本研究基于运动参数想象的实验范式,将速度和力划分为不同的等级,通过识别不同级别的握速和握力想象来增加脑机接口系统的控制模式,为机器人实现方向和速度/力的控制提供一种可能性。3.脑机接口系统中用户快速反应训练的研究。在实时脑机接口系统中,用户需要根据反馈结果及时调整自己的运动想象脑活动,从而来完成特定的任务。在很多情况下,用户很难根据反馈结果快速作出调整,从而产生延迟的控制指令。因此我们需要建立神经反馈系统来对用户进行训练,使其快速地做出反应。我们研究了基于运动相关皮层电位(MRCP)的用户快速反应训练方法,通过识别MRCP信号负峰值的时间延迟可以获得用户的反应时间,通过多次的反馈训练,缩短用户的反应时间。4.基于运动想象脑机接口中多维度控制方法的研究。基于以上研究,我们构建了基于左右手运动想象的虚拟无人机控制系统,传统的无人机控制中只能控制无人机的运动方向。在我们的研究中,除了将左右手运动想象转化为无人机的运动方向之外,还构建了抓握速度和力与脑信号之间的数学模型,通过解调测量到的脑信号,将抓握速度映射为无人机的旋转速度,实现对无人机方向和速度的多维控制。综上所述,我们研究了基于EEG和fNIRS的联合测量技术,实现了对大脑时-空特征的提取,从而提高意图识别率;研究了基于不同级别抓握速度和力的运动想象,来实现多维度、复杂的控制;研究了脑机接口系统中用户的快速反应训练,缩短用户的反应时间,提高脑机接口系统的性能。我们的研究也为脑机接口技术走出实验室,在脑控外骨骼机器人实现多维、复杂控制上的应用作出了进一步的贡献。
其他摘要Brain-computer interface (BCI) technology is a new human-computer interaction way, this technique utilizes brain motor imagery to control peripheral devices directly without normal neural pathways. This is called as “Mind Control”. EEG is a widely used modality in BCI, there is a good time resolution for EEG; however, the spatial resolution is very poor due to the signal overlaping of different brain cortex. fNIRS is similar with fMRI which possess a higher spatial resolution, but low time resolution due to the inherent time delay. The disadvantages of two modalities limit the analysis of brain time-spatial features. Moreover, the control pattern of motor imagery based BCI is simple, only exists in direction. The pattern can’t be used to control the speed or force of the robots. There is a limitation in the application of multi-dimension and complex control. Finally, the users’ operation also influences the BCI performance. The users can not adjust their brain activities quickly according to the feedback, the control results are bad due to the delay command. To solve the existing problems, we performed the following researches: 1. Research on improving the classification accuracy of motor imagery based BCI. EEG is a widely used modality in BCI, there is a good time resolution for EEG; however, the spatial resolution is very poor due to the signal overlaping of different brain cortex. fNIRS is similar with fMRI which possess a higher spatial resolution, but low time resolution due to the inherent time delay. The combination of EEG and fNIRS can remedy their own disadvantages. We researched a hybrid BCI system based on EEG and fNIRS and established our multi-channel fNIRS acqusition system. The results show that the classification accuracy of hybrid modality is higher than single EEG or fNIRS. 2. Research on increasing the number of control pattern in motor imagery based BCI system. The traditional motor imagery based BCI system can only control the robot’s direction, there is a limitation in control pattern. In this thesis, the motor parameters imagery based experimental paradigm is provided. We divided hand-clenching speed and force into different levels, the number of BCI control command is increased by classifying different levels of force and speed. Our research provides a possibility for robot control in direction and speed or force. 3. Research on training quick reaction of the BCI users. The BCI users should adjust their brain activities according to the results of feedback when they are performing the real-time BCI training. In the original stage, it is a little difficult for users to have a quick reaction. Consequently, we should establish a neurofeedback system to train the users. Firstly, we researched the users’ quick reaction training based on movement related cortical potential (MRCP). The users’ reaction time can be calculated from the time latency of negative peak. In this thesis, we designed a neurofeedback system for training users’ reaction. In this system, the reaction time calculated by MRCP is presented for the user in real-time to make them react quickly in the following trials. 4. Research on multi-dimension control in motor imagery based BCI system. Based on above researches, we investigated the control of virtual drone based on left and right hand motor imagery. In traditional research, only the control of direction is investigated. In our research, we decoded left and right hand imagery to control the direction of rotation. And the mathematical model between hand-clenching speed and brain signals was built, the brain signals were decoded and transformed to virtual drone’s rotation speed in real-time. The virtual drone can be controlled in a multi-dimension way. In summary, we investigated the hybrid EEG and fNIRS measurements, to extract the time-spatial features and improve the decoding accuracy. We also researched the motor imagery based on different levels of hand-clenching speed and force, to achieve the multi-dimension and complex control. We further researched the users’ quick reaction training in BCI system to improve the performance. Our research can contribute to the application of brain controlled robot in complex control.
语种中文
产权排序1
文献类型学位论文
条目标识符http://ir.sia.cn/handle/173321/19449
专题工业控制网络与系统研究室
作者单位1.中国科学院沈阳自动化研究所
2.中国科学院大学
推荐引用方式
GB/T 7714
尹旭贤. 基于多模脑机接口的神经信息处理和神经反馈方法的研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2016.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
基于多模脑机接口的神经信息处理和神经反馈(6063KB)学位论文 开放获取CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[尹旭贤]的文章
百度学术
百度学术中相似的文章
[尹旭贤]的文章
必应学术
必应学术中相似的文章
[尹旭贤]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。