中国科学院沈阳自动化研究所机构知识库
Advanced  
SIA OpenIR  > 机器人学研究室  > 学位论文
题名: 基于脑认知的脑控机器人接口研究
其他题名: Brain-controlled Robot Interface Based on Brain cognition
作者: 伏云发
导师: 李洪谊
分类号: TP242
关键词: 脑控机器人接口 ; 脑认知 ; 运动速度想象 ; 握力变化想象 ; 脑-机接口
索取号: TP242/F81/2012
学位专业: 机械电子工程
学位类别: 博士
答辩日期: 2012-05-24
授予单位: 中国科学院沈阳自动化研究所
学位授予地点: 中国科学院沈阳自动化研究所
作者部门: 机器人学研究室
中文摘要: 基于脑认知的脑控机器人接口(Brain-controlled robot interface, BCRI)是一种新型的人-机器人接口技术,是脑-机器接口/脑-计算机接口(Brain-machine interface, BMI/ Brain-computer interface,BCI)在机器人控制领域的重要应用和研究方向。研究者相继在Nature 、Science和其它重要国际期刊上报道了相关的实验研究和开发,目前已成为国际前沿研究热点。该研究从国防军事战略目的扩展到民用目的,主要用于服务机器人或康复机器人,以开启或增强严重运动残疾人控制机器人或外部设备的能力,从而改善其生活质量。此外,还期望为正常人提供特殊情况下控制外部设备或机器人的能力,也可望进一步提高生活质量。已有若干研究用BCRI控制移动机器人、智能机器人轮椅、机械手、智能车辆、仿人机器人等外部设备。因此对BCRI的研究具有重要的现实意义。然而,由于目前人类对脑的感觉、知觉和认知神经机制的认识和理解还相当有限,这给BCRI的研究带来了巨大的挑战,因此仍然需要系统地深入研究与BCRI密切相关的认知事件脑功能神经机制。依赖有限几种具有局限性的测量手段检测反映脑活动的信号,绕过外周神经和肌肉活动直接控制机器人,需要根据认知事件脑功能神经机制量化计算提取反映认知活动的脑信号特征。此外,目前BCRI还不能完全正确、安全、可靠地实现对机器人的连续、精细和复杂的控制,需要研究思维意图直接控制机器人的运动参数,例如速度和操作力等参数。迄今为止,BCRI系统还缺乏相对典型或一般的数学描述,不利于跨学科研究;也没有相对完善的系统结构,不利于有效实现;没有相对系统地阐述BCRI基本原理和方法,为今后进一步深入地研究,需要做这些研究工作。本文在对国内外文献调研与综述的基础上,以一类非常重要的BCRI—基于MI(Motor imagery)的BCRI为研究对象,针对上述问题进行了研究,也设计了实验研究范式并实施了实验,开展了离线的研究,具体如下:    (1)根据BCRI系统中信息的流向和变换,给出了BCRI一种相对典型的数学描述;通过研究比较已有BCRI系统结构,指出存在的问题,提出新的相对完善的系统结构;围绕BCRI中的控制策略、BMI/BCI模块与机器人多层控制模块的适应和融合、BCRI中的脑信号自适应分类算法以及人、BMI/BCI模块和机器人控制系统的三边自适应探讨了BCRI基本方法;对涉及的基本原理进行阐述,有利于跨学科研究。    (2)为今后进一步深入地研究基于MI和EEG (Electroencephalogram)的BCRI,探讨了MI的神经电生理机制—ERD/ERS(Event-related desynchronization/ event-related synchronization), 完整地提出了基于EEG运动想象ERD/ERS时-空-频特征研究准则。基于一种量化计算模型的结构、功能定位以及传感电极定位方法,讨论了MI的神经电生理参数量化计算模型;    (3)基于提出的新的运动速度想象研究范式,探索对运动速度想象具有反应性的脑电节律活动并进行单次识别,正确识别率达到70%以上,具有实用的价值,与Gu等对想象速度的识别率相比具有可比性(他们取得的错误分类率是30.4±3.5%)。结果表明,尽管运动速度想象的单次识别是一个困难的挑战,但通过精心设计研究范式,适当训练被试,能够诱发出对速度起反应的特征频带,基于脑电单次识别运动速度想象是可行的,该研究可望能够为BCRI提供额外的新的速度控制参数;    (4)基于提出的新的左、右手握力或想象握力研究范式,探索左、右手握力运动相关皮层电位(MRCP)的时域特征表示和单次解码握力运动参数。本研究进一步证实了MRCP可以表征运动规划、运动执行和运动监控的脑神经机制过程,这意味着基于EEG解码握力运动相关参数可以考虑把MRCP作为特征。基于MRCP解码左、右手大、小两种握力变化模式,11个被试平均最小误分类率为20±5%,与do Nascimento OF等人基于MRCP识别想象等距足底弯曲扭矩生成速率取得的平均最小错误分类率(17.4 ± 8.4%)具有一定的可比性,也与Gu Y等人基于MRCP在中等扭矩生成速率下识别两种目标扭矩(30 %和 60% 最大自愿收缩扭矩)取得的平均最小错误分类率(26 ± 13%)具有一定的可比性。这些表明基于运动相关皮层电位单次解码左、右手握力运动参数是可能和可行的。与传统的仅识别参与运动或想象运动的肢体类型以及解码单侧肢体运动或想象运动参数的研究相比,本研究可望能够为BCRI提供增加的新的力控制参数。    本文的工作可望为BCRI的进一步深入研究和开发工作起到促进作用,同时架起BMI/BCI与机器人控制领域的桥梁,加强多学科协作研究的局面。在本研究的基础上,完善提出的运动速度想象和握力变化想象实验研究范式,相信未来深入的离线研究和在线研究将为BCRI提供精细的运动参数指令。
英文摘要: Brain-controlled robot interface (BCRI) based on brain cognition is a new type of human-robot interface technology which is an important application and research direction for brain-machine interface (BMI) / brain-computer interface (BCI) applied in robot control field. Many experimental researches and developments for BCRI were reported by Nature, Science and other important international journals and it has become an international frontier research hotspot. The study was expanded from the   military strategic objective for the national defense into civil purpose which was mainly used in the service robots or rehabilitation robots to open or enhance severe motor disabled patients' ability to control robots or external devices, so as to improve the quality of their lives. In addition, it is also expected to provide ability for healthy persons to control external devices under special circumstances and is also expected to further improve the quality of their lives. BCRI has been used to control devices such as mobile robot, intelligent robot wheelchair, manipulator, intelligent vehicle, and humanoid robot in the past and present studies. Therefore, research on BCRI has an important practical significance. However, because our understanding for the neural mechanism of feeling, perception and cognition is still very limited, this situation brings huge challenges to BCRI research. Therefore, there still need a system and in-depth research on brain functional neural mechanism of cognitive events closely related to BCRI. Direct brain-controlled robot by bypassing brain peripheral nerve and muscle with signals reflecting brain activity detected by several measuring methods having limitations needs quantitative calculation and extraction for brain signal features according to brain functional neural mechanisms related to cognitive events. In addition, at present BCRI still can't completely realize continuous, fine and complex control of robot with high safety and reliability. It needs to research on direct control of the robot movement parameters such as speed and force parameters by thinking intentions. So far, BCRI system also are shortage of relatively typical or general mathematics description which goes against interdisciplinary research; and no relatively perfect system structure is conducive to effective implementation; no relatively systematically expounding the basic principles and methods for BCRI. In order to further explore BCRI in the future, it needs to research on these problems. By  my investigating and reviewing home and abroad literatures, and aiming at an very important BCRI based on motor imagery (MI) , this paper studies the above problems and also designs experimental research paradigms and carries out these experiments and offline studies, specific as follows:    (1) According to the information flow direction and transformation in BCRI system, a relatively typical mathematics description for BCRI is presented; By  studying and comparing existing BCRI system structure and after pointing out the existing problems, a new relatively perfect system structure are proposed; The basic methods for BCRI including the control strategies, adaptation and fusion between BMI/BCI module and the robot multilayer control architecture (RMCA), adaptive feature extraction and adaptive classification algorithm for brain signal and the trilateral adaptive method among human, BMI/BCI module and RMCA in BCRI are discussed; The basic principles related to BCRI are analyzed and helpful for BCRI interdisciplinary research.    (2) To further study BCRI based on MI and EEG in the future, the neural electrophysiology mechanism for MI: ERD/ERS is discussed and time-spatial- frequency features study criteria for ERD/ERS related to MI based on EEG are fully put forward. Based on a selected brain structure, functional localization and electrode positioning method for quantitative calculation, a quantitative calculation model for MI nerve electrophysiology parameters are discussed.    (3) A new paradigm on imagined movement speeds was presented in the study. EEG rhythmic activities reactive to imagined movement speed were explored and identification of single-trial EEG related to imagined movement speed was investigated under the new paradigm. The accurate recognition rates were above 70% which had practical value and comparable with the recognition rate of imagined speed achieved by Gu Y. etc. (their error rate was 30.4±3.5%).The results show that although single-trial identification for imagined movement speeds based on EEG is a challenging research, distinctive frequency band activities related to imagined speeds can be evoked by carefully designing research paradigms and properly training subjects and single-trial identification of imagined movement speeds based on EEG is feasible.The study is expected to provide additional new speed control parameters for brain-computer interface.    (4) Based on a new research paradigm for grip force movement or imagined grip force movement involved left and right hands, the study explores time-domain feature representation for grip force movement-related cortical potentials (MRCP) of left and right hands and the decoding of grip force parameters based on a single trial EEG activity. The study further demonstrates that MRCP may characterize brain neural mechanism process for planning, execution and precision of a given motor task and means that features related to MRCP can be used for the decoding of grip force parameters based on EEG. Decoding of two modes of grip force variation involving left and right hands based on MRCP, the average minimum misclassification rate across 11 subjects was (mean ± SD ) 20±5% which was comparable with the average misclassification rate (17.4 ± 8.4%) for discrimination of imaginary RTD (rate of torque development) of isometric plantar-flexions based on MRCP by do Nascimento OF and also the average misclassification rate(26 ± 13%)for identification of  the two target torques (30 and 60% of the maximal voluntary contraction torque) under  moderate RTD based on MRCP by Gu Y. It is feasible for the single-trial decoding of grip force parameters based on MRCP. Compared with the traditionally only identifying limb types involved in movement or imagined movement and the decoding of their parameters involving unilateral limb, the study may provide some additional and fine control instructions for BCRI/ BMI / BCI to achieve a more complex control.     The study is expected to play a role in promoting further research and development for BCRI and bridge the gap between BMI/BCI and robot control to strengthen the multidisciplinary cooperation study. On the basis of this study, by further improving the proposed experiment research paradigms for imagined movement speed and imagined grip force variation, I believe that the in-depth offline research and online research in the future can provide fine motion parameters instructions for BCRI.
语种: 中文
产权排序: 1
内容类型: 学位论文
URI标识: http://ir.sia.cn/handle/173321/9386
Appears in Collections:机器人学研究室_学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
基于脑认知的脑控机器人接口研究.pdf(2802KB)----限制开放 联系获取全文

Recommended Citation:
伏云发.基于脑认知的脑控机器人接口研究.[博士学位论文].中国科学院沈阳自动化研究所.2012
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[伏云发]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[伏云发]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.

 

 

Valid XHTML 1.0!
Copyright © 2007-2016  中国科学院沈阳自动化研究所 - Feedback
Powered by CSpace