【作者】 张林；

【导师】 刘蓉；

【作者基本信息】 大连理工大学 ， 生物医学工程， 2014， 硕士

【摘要】 脑机接口(BCI)是一种不依赖于神经肌肉组织的新型人机交互工具,目前在康复医疗、生物医学、娱乐等领域都具有广阔的发展前景,脑控机器人系统就室一个重要的研究领域。脑控机器系统在BCI技术的基础上结合机器人技术,实现大脑意图对机器人的控制,从而为危险作业、残疾人辅助设备提供了新的实现途径。但目前异步BCI脑电识别准确率不高,难以精准控制机器人的运行,因此需要机器人智能技术的辅助,从而实现共享控制。共享控制的引入使得脑控机器人系统的性能有了很大的改善,但当前的共享控制方法大多是简单的结合智能控制,并没有从控制方法的角度L考虑如何实现人和机器人的协调控制,缺乏对错误脑控命令的修正与补偿。其次,在系统运行期间,难免会遇到突发事件以及不确定因素影响传感器精度下降等问题,这会对整个系统的稳定性造成很大影响。另外,当前共享控制方法大多采用离散事件系统模型(DES),其异步特性可以很好的描述脑控机器人系统,但由于同一时刻只能确定唯一状态,当机器人住处于两种或者几种行为状态临界条件时,就会出现频繁的状态转换,从而影响系统的稳定性。本文基于以上存在的问题,提出了基于模糊离散事件系统(FDES)的共享控制方法,通过模糊状态的行为融合降低了机器人由于临界状态转换造成的不稳定性,而模糊属性则增强了系统的适应性。另外,FDES的模糊监督方法对人的控制命令起大到了很好的补偿作用。其次,本文在对BCI异步竞赛数据分析的基础上,采用了离散小波变换和CSP进行特征提取,结合SVM和概率统计的方法实现了对无休息状态训练样本的异步数据分类,分类准确率较冠军组算法提高了13%。最后,通过结合脑电识别算法和共享控制方法建立了由BCI2000和MobileSim机器仿真软件构成的脑控机器人系统,并在线验证了实验效果,对比基于FDES和基于DES的共享控制方法,受试者1的完成任务次数提高了100%,平均完成时间减少了7%,平均碰撞次数减少了50%,平均脑控命令次数减少了2%；受试者2的平均完成时间减少了22.3%,脑控命令次数减少了19.5%。另外,NASA-TLX量表的结果也表明基于FDES的共享控制方法能够减少疲劳程度。实验结果表明了本文的共享控制方法的有效性。更多还原

【Abstract】 Brain computer interface (BCI) is a new tool of human-computer interface which do not depend on the neuromuscular organization, it has a broad prospects in the field of medical rehabilitation, biomedical, entertainment, etc. Related research groups have increased to a large number, the research value of BCI has received wide attention, brain-actuated robot is one of the most important research fields. Brain-actuated robot can control the robot with brain signal based on ne BCI technology and robot technology, therefore dangerous working and assistive devices can be achieved with a new method. But due to the low accuracy rate of brain signal recognition in asynchronous BCI, robot cannot be controlled exactly. In order to solve this issue, shared control with assistant technology of robot should be introduced.Shared control enhances the performance of brain actuated-robot greatly, but the current shared control method only combines intelligent control simply, coordination control of operator and robot is not considered in aspect of control method. Furthermore, the stability of system may be influenced by accidental event and uncertain factors. In addition, the current shared control methods use discrete event system (DES) to model, the asynchronous nature of DES is suitable for system description. But the stability of system will be affected by the chattering effects which will be introduced into the system when switches between the hard boundaries of multi-states.To analyze the drawback of shared control, the method based on fuzzy discrete event system (FDES) is proposed in this paper. This method can reduce the instability of system caused by switches between hard boundaries of states and increase flexibility of system. In addition, the fuzzy supervisory method can compensate the brain control commonds. Moreover, wavelet transform and common spatial pattern (CSP) are used for feature extraction, support vector machines (SVM) and probability statistics are used for classification of asynchronous data without relax training data, the classification accuracy increase13%than champion algorithm. Finally, brain actuated robot system is composed of BCI2000and MobileSim based on algorithm of brain signal recognition and shared control. To compare with shared control of FDES and DES, subject1’average completion times enhance100%, average completion time reduce7%, average collision times reduce50%, average brain control commands reduce2%; Subject2’average completion time reduces22.3%, brain control commands reduce19.5%.Shared control based FDES also can reduce fatigue according to the results of NASA-TLX. The result of online experiment verify the effectiveness of shared control method.更多还原