CASIA OpenIR  > 毕业生  > 硕士学位论文
基于精细运动想象范式的机械臂控制系统设计与研究
王宇
2024-05-13
Pages85
Subtype硕士
Abstract

-机接口技术旨在建立起大脑和外部环境设备的双向信息通路,通过对大脑信号的解码以及反馈刺激,在不依赖于外周神经组织和肌肉组织的条件下完成指定任务或修复神经功能。其中,基于运动想象的脑-机接口系统在康复医疗和运动能力增强方面有着广泛的应用前景。在对外部设备进行精细控制的需求中,相较于可能导致运动意图与实际设备执行指令不匹配的简单肢体想象任务,精细运动想象任务由于其在提升设备控制精度和扩大可执行指令集方面的显著潜能,已成为运动想象研究领域的关键研究焦点。

本文致力于精细运动想象脑-机接口的机械臂控制系统的设计与优化,论文的主要工作和创新点归纳如下:

1.   本研究搭建了一套基于精细运动想象脑-机接口的机械臂控制系统。该系统通过解码单侧肢体不同关节的精细运动想象脑电图模式,并结合咬牙的脑电信号检测,实现了二维平面内对目标物体的抓取。系统分为三个主要部分:数据采集平台、数据预处理与算法平台、机械臂控制系统。数据采集平台实现对脑电信号的收集,并配备了视觉及听觉刺激模块,以满足不同模式下脑电数据采集需求;采集到的脑电信号通过TCP/IP协议传输至数据预处理与算法处理平台,该平台进行实时预处理与解码,并将解码结果传递至机械臂控制系统中;机械臂控制系统根据接收到的解码结果,执行预设动作,从而实现对机械臂的控制。该系统不仅为运动障碍患者提供了一种新型交互方法,而且可以为中风康复患者提供更加精细的康复训练方式。

2.   为了实现使用者对外设控制系统的高效控制,本研究基于自主搭建的基于精细运动想象脑-机接口的机械臂控制系统,设计了一套单侧肢体精细运动想象任务的多阶段训练范式,并采集了共计22名被试,66人次的实验数据。该范式由视觉反馈训练和三阶段的控制实验组成,首先,通过视觉反馈训练增强被试在精细运动想象的脑电信号可区分性,然后,三个渐进式的控制实验帮助被试通过想象精细运动和执行咬牙来控制机械臂。实验结果显示,视觉反馈训练实验有效地增强了被试在精细运动想象任务中的信号可区分度(p<0.1)。此外,在控制实验中,被试操作机械臂完成任务的效率也得到了显著提高(p<0.1),该范式为精细运动想象脑-机接口系统的使用者训练提供了指导。

3.   针对单侧肢体精细运动想象任务解码中准确率较低的问题,本研究提出了基于多频带黎曼特征与卷积神经网络联合的解码方法。该方法利用脑电数据的时频空特性,设计多频带黎曼流形进行特征提取,并且基于卷积神经网络的特征提取网络进一步挖掘高维特征,进行精细运动想象任务解码。在对自采的单侧肢体精细运动想象实验数据进行离线及在线测试中,我们的模型在精细肢体运动想象的解码任务中显示出优异的性能,不仅离线准确率达到了84.38%,在两组控制实验中也分别获得了74.54%76.59%的准确率,显著优于传统方法,为推动精细运动想象脑-机接口系统的实际应用提供了技术支持。

本文面向精细运动想象脑-机接口中,系统集成与测试验证方面的研究匮乏的问题,搭建了一套基于精细运动想象脑-机接口的机械臂控制系统,设计了多阶段训练范式,采集和分析了脑电和任务数据,并设计了精细运动想象解码算法,缓解了想象任务与外部设备输出模式之间的不一致性问题,为进一步拓展运动想象脑-机接口在康复治疗领域的应用提供关键的技术支撑。

Other Abstract

Brain-computer interface (BCI) technology aims to establish a bidirectional information pathway between the brain and external environmental devices. By decoding brain signals and providing feedback stimulation, it accomplishes specified tasks or repairs neural functions without relying on peripheral nervous and muscular tissues. Among these, BCIs based on motor imagery have broad application prospects in rehabilitation medicine and the enhancement of motor abilities. In the demand for precise control over external devices, fine motor imagery tasks have emerged as a critical research focus within the field of motor imagery. These tasks are preferred over simpler limb imagery tasks, which may lead to discrepancies between intended movements and actual device commands, due to their significant potential in enhancing device control precision and expanding the operational command set.

This paper is dedicated to the design and optimization of a robotic arm control system for fine motor imagery BCIs, the main work and innovations of this thesis are summarized as follows:

1.This study has constructed a robotic arm control system based on fine motor imagery BCIs. The system decodes the electroencephalogram (EEG) patterns of fine motor imagery from unilateral limb joints, combined with the detection of teeth-clenching EEG signals, to achieve the grasping of target objects in a two-dimensional plane. The system consists of three main parts: a data collection platform, a data preprocessing and algorithm platform, and a robotic arm control system. The data collection platform collects EEG signals and is equipped with visual and auditory stimulation modules to meet the needs of EEG data collection under different modes; collected EEG signals are transmitted to the data preprocessing and algorithm platform via TCP/IP protocol. This platform performs real-time preprocessing and decoding, and the decoded results are passed to the robotic arm control system, which executes preset actions to control the robotic arm. This system not only provides a new interactive method for patients with motor disabilities but also offers more refined rehabilitation training for stroke recovery patients.

2.To achieve efficient control over peripheral device systems by users, this study developed a multi-stage training paradigm for unilateral limb fine motor imagery tasks, based on a self-constructed robotic arm control system utilizing fine motor imagery BCI. Data were collected from a total of 22 subjects, amounting to 66 experimental sessions. The paradigm consists of visual feedback training and three stages of control experiments. Initially, visual feedback training was used to enhance the distinguishability of EEG signals during fine motor imagery. Subsequently, three progressive control experiments aided subjects in controlling the robotic arm through the imagination of fine movements and the execution of teeth clenching. The experimental results demonstrated that the visual feedback training increased the signal distinguishability in fine motor imagery tasks (p<0.1). Furthermore, in the control experiments, the efficiency with which subjects operated the robotic arm to complete tasks significantly improved. This paradigm provides guidance for training users of fine motor imagery BCI systems.

3.To address the issue of low accuracy in decoding fine motor imagery tasks of unilateral limbs, this study proposed a decoding method combining multi-band Riemannian features with convolutional neural networks. Utilizing the time-frequency-spatial characteristics of EEG data, this method designs a multi-band Riemannian manifold for feature extraction. Additionally, a feature extraction network based on convolutional neural networks further mines high-dimensional features for decoding fine motor imagery tasks. Our model demonstrated excellent performance in offline and online tests of self-collected fine motor imagery data from unilateral limbs, not only achieving an offline accuracy rate of 84.38% but also obtaining 74.54% and 76.59% accuracy rates in two control experiments, significantly surpassing traditional methods and providing technical support for the practical application of fine motor imagery BCI systems.

This paper addresses the lack of research on system integration and testing verification in fine motor imagery brain-computer interfaces (BCI). We constructed a robotic arm control system based on fine motor imagery BCI, designed a multi-stage training paradigm, collected and analyzed EEG and task data, and developed a fine motor imagery decoding algorithm. This work alleviates the inconsistency between imagined tasks and the output modes of external devices, providing crucial technical support for further expanding the application of motor imagery BCI in the field of rehabilitation therapy.

Keyword脑-机接口 运动想象 机械臂控制系统 单侧肢体 机器学习
Subject Area人工智能其他学科
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/56487
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
王宇. 基于精细运动想象范式的机械臂控制系统设计与研究[D],2024.
Files in This Item:
File Name/Size DocType Version Access License
(明版)基于精细运动想象范式的机械臂控制(3839KB)学位论文 限制开放CC BY-NC-SA
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[王宇]'s Articles
Baidu academic
Similar articles in Baidu academic
[王宇]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[王宇]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

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