ArticlePDF Available
科技导报 20 21,3919
www.kjdb.org
脑机接口技术发展新趋势
—基于 2019—2020 年研究进展
陈小刚1杨晨 2陈菁3高小榕 3*
摘要 脑机接口旨在通过直接从大脑信号中实时解码用户意图来为辅助设备提供丰富、
大的命令信号。近年来机接口技术的理论和实际应用的研究进展迅速技术日趋成熟
其应用领域 也在不断扩大。概述 了 20192020 年脑机接口领域在硬件、算法、范式、应用等
方面取得的重要研究进展和发生的热点事件,展望了未来脑机接口技术的发展趋势。
关键词 脑机接口应用系统;关键技术
收稿日期:2020-06-08修回日期2020-06-23
基金项目国家重点研发计划项目2017YFB1002505广东省重点领域研发计划项目2018B030339001中央高校基本科研业务费专项资
金项目33320190153332018191
作者简介:陈小刚,副研究员,研究方向为脑机接口,电子信箱:chenxg@bme.cams.cn高小榕通信作者教授,研究方向为脑机接口,电子信
箱:gxr-dea@tsinghua.edu.cn
引用格 式:陈小,,陈菁 ,.脑机接口技术发展新趋势
—基于 20192020 年研究进展 [J]. 科技导报, 2021, 39(19): 56-65; doi:
10.3981/j.issn.1000-7857.2021.19.007
1. 中国医学科学院北京协和医学院生物医学工程研究所,天津 300192
2. 北京邮电大学电子工程学院北京 100876
3. 清华大学医学院生物医学工程系,北京 100084
脑机接口brain-computer interfaceBCI
通过检测用户意图直接控制外部设备,而为恢
感觉和运动功能以及治疗神经系统疾病开启了一
扇全新的门。尽管侵入和非侵入式脑接口
系统在可实现任务复杂性方面有所不同,但两类脑
机接口都能成功展现出辅助恢复肢体功能的能力,
[1-6]
近年来,脑机接口技术不断取得破,也带动了产
业界的进一步投入。例如,Elon Musk 的脑机接口
技术公司 Neuralink 2 019 年发布了脑机接口植入
机器 2019 Facebook 斥资 10 亿美
机接口公司 CTRL-labs2019 亚洲消费电子展
CES Asia 2019上,日产汽车展示了解读大脑信
brain-to-vehicle 术 。在 2020 5
月,米科技和中国科学技术大学先进技术研究
共同建立了“脑机智能联合实验室”展示了国内企
业对这一技术的关注。
本研究从应用系统实现、关键技术发展和未来
发展趋势等方面 回顾 20192020 年的脑机接口领
域的前沿进展。
56
科技导报 20 21,3919 www.kjdb.org
1应用系统实现:交流与控制
1.1 沟通交
语音解码脑机口技术能将神经活
接转换为语音信号,对由于神经功能障碍而无
正常交流的群体具有革命性的意义。2019 4月,
加州大学旧金山分校的研究团队基于脑机接口技
术设计了一种新型的神经解码器,该解码器能够通
过提取大脑皮层活动对发声器官的运动情况来实
现语 音的合 成(图 1[6]
不仅 研究进一
发现便是在受试“默读”在不出声音的
情况下读出句子时,样的解码方法也能够实现
音的合成表明该系统有望潜在应用于不能发出
音的人[6]。这些研究成果展现出通过脑机接口技术
帮助患者恢复口语交流能力的临床可行性并发
Nature上。
2蓝)和答红)过程的实时语音解码示意
1通过皮层脑电合成语
随后的 2019 7月,该研究团队在同样的实验
系统上展示了基于高密度 ECoG 自然
问答对话系统2[ 7]
。利用在对话中记录下的脑
信号,能够确定受试者何时在听、何时在说,并且能
够预测所或所说的是什。由于特定的题只
可能对应特定答案在这一系统中研究者
根据解码后的问题来动态更新答案的先验概率,
而实现了为准确的回答容解码。该系对于
生成语音和感知语音的解码准确率分别高达 61%
76%[7]。这一成果阐明脑机接口技术可以在交互
式对话环境中实时解码语音,于无法交流的患
具有意义该工作发 Nature Communica⁃
tions上。
2020 3月,加州大学旧金山分校的研究团队
利用受试者朗读文本时收集的 ECoG 号,成功训
练了一个可以将 ECoG 到端译”
文字的深度循环神经网络 模型(图 3[8]
通过在侧
裂周围区植入电极并采集 ECoG 号,该研究可以
实现 者当 子的 4位病
内脑电解读错误率最低可以达到 3%[8]该研
果发表在Nature Neuroscience上。
除了上述及的用侵入式机接口实
言交流功能外,非侵入式脑机接口在恢复交流功能
方面也展现出较大的潜力。2019 47
电视台的挑战不可能节目中,清华大学研究团队
展示了一脑机接口打字入系统。在这系统
57
科技导报 20 21,3919
www.kjdb.org
3将皮层脑电转录成文
的帮助下全身只有眼球和嘴角可以活动的渐冻
甲成 并完了与
“合诵”。这套统展示了脑机口系统在帮助
冻症群体重拾交流能力的潜力。
1.2 触觉和运动恢复
除了帮助患者恢复交流能力之外,脑机接口的
另一项经典应用场景则是帮助运动障碍群体
实现运动能的恢复。在国巴特尔纪念究所
合作 的研 基于
术,名脊髓损伤患者利用其手部残存的触摸信
实现了触觉和运动功能的 同时恢复(图 4[9]
在这
项研究中脑机接口系统从初级运动皮层活动反
的运动意图中提取出患者残余的、无法被患者知觉
感知的手部触觉信号,并将该信号进行增强后反馈
给患者,从而实现皮层内控制的环感觉反馈,
可以通过触摸信号调节握力[9]。在这样一套闭环反
馈的脑机接口帮助下,进行康复训练后患者几乎完
全恢复了感知物体触摸的能力,多种感觉运动功能
也得到了显著改善。这些结果表明脑机接口可
从大脑皮层采集低于知觉反应范围的神经信号,
将其转换为有意识的知觉从而显著增强功能。
项工作展示了首个同时恢复运动与触觉功能的脑
机接口系统,并在Cell上发表。
1.3 运动控
法国格勒布尔学及合作究团队报
一项脑控骨骼的研究成。一名四肢瘫患者
在一款处于实验室测试阶段的大脑控制外骨骼系
统帮 实现 行走[ 10]研究在他
的上 肢感觉 运动区植入 了 2个双侧线硬膜
以获ECoG ECoG
信号通过自适应解码算法在线处理以将命令发
至效应器拟化身或外骨2年间人在
中利用虚拟化身模拟行走并在不同的触控任务
手腕旋转过程八自度的方式进行侧、
关节肢的运动功率64.0%在实验室的
骨骼 中的成功 率为 70.9%。这项工作首次成功验
证了长期使用无线硬膜外多通道记录仪的可能,
首次将人类长期临床应用所需的所有技术要(硬
脑膜 记录无线和发 射、多通ECoG
在线解码以及完全嵌入进行了组合。与微电极
列相比,硬膜外 ECoG 侵入性较小,而效果相似
The
Lancet Neurology上。
2020 1月,浙江大学研究团队也实现了国内
第一例植入式脑机接口临床研究。在植入电极后,
患者可以利用大脑运动皮层信号精准控制外部机
械臂与机械手实现三维空间的运动首次证明高
4同时恢复运动与触觉功能的脑机接口系统
58
科技导报 20 21,3919 www.kjdb.org
患者利用植入式脑机接口进行复杂而有效的运动
控制的可行性。
来自卡耐梅隆学和明尼达大学的
人员则为脑机接口的机械臂控制提出了非侵入的
决方 脑电 者可
臂,实 现对连续随机运动目标的实时跟5[11]
该技术通过连续的追踪任务和相关的训练范式增
加用户的参与度显著改善了基于脑电的神经解
效率,并且能够允许用户对机械臂实现高分辨率控
制。该技术可以帮助用户通过脑机接口系统实现
对计算机标的连续跟踪跟随。在该研成果
中,研究人员将计算机光标的连续跟踪提高了
500% 以上,解决了机械臂跟随光标的流畅度问题,
同时结合在线无创神经成像,脑机接口的控制
平提 升了10%[11]。研究人员将该技术应用于现
实任务中发现从控制不受约束的虚拟光标运动
控制机械臂运技术乎可以做到完过渡
这种高质量的神经解码能力与非侵入式机械臂控
制的实际应用相结合,将对利用非侵入式脑机接口
开发和实现神经机器人技术产生重大影响。目前
68 名健康受试者身上进行了测
试,团队还将对患者进行临床试验。这项研究
表了无创脑机接口技术发展的重要一步,有望成为
如智能手机一样帮助每个人的辅助技术。该研究
成果发表在Science Robotics上。
2关键技术进展:硬件新算法、
范式、应用
2.1 新硬件
脑机接口硬件要涉及电和信号采
统。对于侵入式脑机接口而言,需要具有生物相容
性、安全性长期植入的材料特的电极;而非侵
式脑 于舒 便携的
式。脑机接口硬件的创新将极大地推动脑机接口
技术的实用化、产品化。2019 7月,Elon Musk
Neuralink 公司发布了一款可扩展的高带宽脑机接
口系统[12]。该系统包含小而灵活的电线”阵列,
每个阵列 多达 96 根线,每 根线带有 32 个电极共分
布了多达 3072 个电极。该系统还包含一个神经
该机 器人 6 线
192 个电极)每根 线以以 级的
独插入特定大脑靶区中,以避开表面血管。电极阵
列被封装在一个小的可植入设备中该设备包含
于低功耗板上放大和数字化的定制芯片,3072
通道封装所占面积小于 23 mm×18.5 mm×2 mm
6[12]
。一 根 USB-C 电缆可提供设备的全带宽数据
流传输,并同时记录所有通道的数据。相较于传统
的实验室中所设计的样机往往较为简陋,未实现工
程上的充分优化,Neuralink 这一
成化、自动化的脑机接口系统展示了工业界的关注
对脑机接口实用化进程的重大意义。
5用于连续随机目标跟踪的脑机接
控制机械臂系统 6植入大鼠的传感器设
59
科技导报 20 21,3919
www.kjdb.org
日本熊本学和口大学的究团队将
外光谱、皮层脑电和负温度系数热敏电阻传感器的
多通道测功能集成到单设备中。该设使用
柔性印刷电路技术和半导体微制造技术进行制造
和组装,以实现与硬膜下植入兼容的传感器组件的
高密度集成,且具有类似于神经外科手术中硬
下条状电极的紧凑外形,可以提供有关大脑皮层活
动各个方面的益信息,并被证是术前、术中和
术后诊断的有力医学手段[13]
除了上述提及的有创脑信号获取方法外,在无
创脑信号获取硬件方法也取得了进展。英国诺丁汉
大学及合作研究团队开发了一种基于自行车头盔改
造且完全符合生命周期的可穿戴脑磁系统该系
能够为所有年龄段的受试者提供高保真数据
需限制 受试者的活7[14]
因此可以使用单一
系统测量儿童、青少年和成人在外部环境中大脑如
何做出反应并适应自然事件的能力有望提供有
早期生命中枢神经系统功能发育的机制性见解[14]
这项工作揭示了一种功能成像的新方法,为研究健
康和疾病中的神经发育提供一个强大的平台。
实现舒适、便携的信号获取方式是推广非侵入
式脑机接系统的重要前。佐治亚理工院及
合作研究团队道了一个完全便携式线
的头皮电子系统其中包括一组干电极和一个柔
膜电路8[1 5]
。利用卷积神经网络进行时域分析
可对稳态视觉诱发电位进行准确、实时的分类。相
比于商用系统,柔性电子产品因显著降低噪声和电
磁干扰能提高诱发电位测性能。两通的头
皮电子系统获得了 122.1 bit/min 信息传输率,
许对电动轮椅电动汽车和无键演示进行无线
实时 用的 [15]麦奥斯大
合作研究团队开发了一种基于干式接触电极的外
耳道脑电采集系统,系统包括嵌入个性化软耳
中的主动屏蔽和纳米结构电极。通过 12 名受试者
和听觉稳态响应、稳态视觉诱发电位、失匹配负波、
alpha 调制 4范式的验证,所开发系统的性能与靠
近耳 头皮 性能[16]。清华大学及合
作研究团队开了一种高成本率、易于制造、
活、鲁棒且凝胶的银纳米线/聚乙稀醇缩丁醛/
聚氰胺海的脑电电极。于银纳米线的面金
属化海绵具有高电导率,而重量却保持不变。柔
软的海绵架构和自锁式银纳米线结合在一起
极提供了卓越的机械稳定性和绕过头发的能力。
利用该电极放置在无毛皮肤时,所构建的稳态视觉
诱发电位脑-机接口的性能与导电膏支持的常规
电极的脑机接口性能相当。尤为重要的是该电
在毛 的性 显著[17]。这一成果显示
出该新型电极有望替代脑电采集的常规电极。另
外,电数据处理芯片化有望为脑机接口技术走
民用化、便携化、可穿戴化及简单易用化开辟道路。
2019 年,天津大学和中国电子信息产业集团在第
三届世界智能大会上发布了脑机接口专用芯
语者”
2.2 新算法
对于侵入式脑机接口而言,脑机接口技术临床
应用的关键障碍是植入皮层电极所记录的神经活
动会随时变化。而神经录的不稳定性导致
临床脑机口无法控制。内基梅隆大学合作
研究团队利用低维神经流即描述神经元之间
定相关模式的维空对齐开发了一种基于
流形的神经信号稳定器,以实现脑机接口信号的稳
7装在经过改装的自行车头盔中的
光泵磁力仪测量脑磁信号
8具有完全便携式和无线的头皮电子设备
60
科技导报 20 21,3919 www.kjdb.org
定输入,从而在神经记录不稳定时依旧能够维持脑
机接 [18]。斯坦福大学及合作研究团队提出
了 一 种 时 间 约 束 的 稀 疏 组 空 间 模 式temporally
constrained sparse group spatial patterns通过同
优化共空间模式中滤波器频带和时间窗长实现
一步 想象 机接性能[19]。英国埃塞
克斯大学及合作研究团队提出了一种用于脑机音
乐接brain computer music interfaceBCMI
性化情感态检测方法。比于基于群体检测
方 法population-based detection method个 性 化
情感状态检测方法的正确率更高,情感效价和唤醒
度的 平均正确 率分别提 高了 10.2% 9.3%[20]
国高丽大学和新加坡南洋理工大学的研究团队提
出了一种基于受试者的分段频谱滤波subject-de⁃
pendent and section-wise spectral filteringSSSF
法,在从预处理技术频谱滤波角度提高
movement-related cortical poten
tialMRCP的解码性能。该方法考虑了两个不同
的时间段的受试者个性化 MRCP 频谱频率特征,
单试次 MRCP 检测实现了 0. 86 的平均解码性能,
证了所提出的 SSSF 方法比常规方法可以包含更多
有意义的 MRCP [21] 。华中科技大学的研究团
队提了一新颖流形入知 迁移manifold
embedded knowledge transferMEKT方法。该方法
首先在黎曼流形中对齐 EEG 试次的协方差矩阵,
提取切线空间中的特征;然后通过最小化源域和目
标域之间的联合概率分布偏移来实现域自适应,
时保留其何结构。该方可以处理一个多个
源域,并且可以高效地进行计算。针对于大量源域
的情况,该团队还提出了域迁移性估计domain
transferability estimationDTE方法以识别有利
的源域 。对来自 2个不 同 BCI 范式的4个脑电数据
集进行的实验表明,MEKT 优于几种最先进的转移
学习方法,并且当源受试者的数量很大时,DTE
以减少一半以上的计算成本,几乎不会牺牲分
精度[22] 。法国 Aramis project-team 合作研究团队
提出了一种融合方法,该方法能够整合来自同步脑
电和脑磁信号的信息,以提高基于运动想象脑机接
口的分类能。该方法的心在于能够自加权
种模 以优与单
比,于脑电和脑磁的多模态信息能够显著提高
机接口的分类性能[23]
2.3 新范式
斯坦福大及合研究团队用神经元
在四肢瘫痪患上研究了面部头部、手臂和腿部
运动如何在前动皮结区中表征。们发
现上述所有运动在“手结区”均具有较好的表征,
且存在着四肢联系起来神经编码。这联系
可能有助于大脑将其某一肢体学会的技能转移至
另一肢体中。基于上述发现,研究团队设计一
脑机接口系统,能够利用手结区”的信号精确地解
码四 运动[24]。先前研究者往往认为要控制身
体的不同部位就需要在大脑的多个区域植入电极,
而这一研究成果展示只在一个区域放置植入电极,
就可能实现全身的运动控制,一成果将大大拓
颅内脑机口的应用空间匹兹堡大学的究团
队提出了一种基于运动想象的混合脑机接口
用脑电图记录脑电活动以及利用功能性经颅多普
勒 超 声functional transcranial Doppler ultrasound
fTCD测 量 脑 血 流 速 度 。 研 究 人 员 计 算 了 来 自
EEG fTCD 信号的功率谱特征,并利用互信息和
线性 向量 特征和分 [25] 。与
的基于 EEG fNIRS 的混合脑机接口相比,所构建
的系统能够以较短的任务持续时间实现相似或更
高的准确率。多伦多大学及合作研究团队利用近
红外光谱成像技术实现了在线三分类想象言语
imagined speech
默念短语“是”“否”来直接回答是或否问题该脑
机接口 还能识别无限制 休息状态而构成 了 3
可识别 的任务。在最后 3在线实验中,所有受试
者的 平均在线 分类正确 率达 64.1%
明,想象言可以用作选定用户可靠激活任务
以开 为直 机接[26] 。对象运
想象 action
observationAO有助于检测用户的运动意图。东
京农工大学的研究团队研究了动作观察的目标对
参与者或他人的手)是否响想象运动时
的大脑活动。研究人员发现,象运动期间动作
61
科技导报 20 21,3919
www.kjdb.org
MI+ownAO所诱发的
感觉运动区 alpha 节律的事件相关去同步强于仅想
象运MI和想象运动期间动作观察来自于其他
MI+otherAO2
具有动作观察的闭环脑机接口设计中应该使用用
户自己肢体的视频[27]
2.4 新应用
脑机接口术除在恢复感和运动功
及治疗神经系统疾病等方面可以发挥作用也已
开始在其领域发挥价值唤醒程度会影个体
的决 判断行为 Yerkes-Dodson 定律指出唤
醒程度与任务执行之间的关系是倒 U形曲线关系,
存在一种对于特定任务的行为执行最佳的唤醒状
态。来自哥伦比亚大学的研究团队验证了可以使
用基于脑电的反馈来改变个体的唤醒程度从而使
他们的任表现显著提高这项工作展示一个
闭环的脑机接口该系统基于脑电信号解码器输
动态 地调bound⁃
ary-avoidance 并根Yerkes-
Dodson 定律提高任务执行效率[28]。该方法有望应
用于不同的任务或用于将自我调节作为目标治疗
的临 床应用来自罗斯 Neurobotics 和莫斯科物
理技术学院的研究团队介绍了另外一种新颖的闭
环脑 口系 [2 9]。该系统可利用受试者的脑电
特征实时重建受试者观察到的或想象的刺激图像,
并将重建的图像作为视觉反馈呈现给受试者。所
提出的技术可以通过将原始刺激替换为受试者的
意念驱动图像重建模型,从而有望用于训练脑机接
口的新用户。脑活动除了可以反映个体的意图和
状态可以体现个体的特质。2019 国科
院半导体研究所的研究团队利用编码调制的视觉
诱发电位实现了一套个体身份识别系统。该系
25 名受试者的个体内和跨个体识别中均获得了
较高的正确识别率。在个体内情况,使用 5.25 s
电数据可获得 100% 的最佳识别性跨个体情
使用 10.5 s 脑电数据可获得 99.43% 的个体区
分效果[30]。该方法有望为个体身份识别提供基于
脑电的解方案。脑机接也为沉浸式虚现实
环境中的通信和控制提供了潜在的工具。中国科
学院半导体研究所和清华大学的研究团队利用
room-scale 虚拟现实头盔开发一个便携式稳态视
觉诱发电脑机接口。通解决虚拟现实刺激
呈现和基于移动脑电信号进行目标识别的问题,
证了脑机接口在移动虚拟现实环境中的应用潜力,
并为利用移动虚拟现实系统开发实用脑机接口提
供了 和方 [31]中国科学
物医学工程研究所及合作研究团队则将脑机接口
技术引入字符号转换测digit symbol substitu⁃
tion testDSST 了基 稳态 诱发
位脑机接口的数字符号转换测试系统[32]
3发展趋势与展望
3.1 高性能脑机接口
尽管近年脑机口在性能获得了较
提高但相比于自然的人机交互目前脑机接口的
通信速 率仍较低进一步 大幅提高。图 9显示
各种 的通 [33]他人互方
比,通信速率仍然是目前限制脑机接口应用的
大障碍。通过脑信号解码技术大幅提高脑机接口
的通信速率,大脑与机器之间建立高效的信息
流通道,是实现高性能脑机接口关键。目前,
何使用先进的算法与大脑进行交互已经引起脑机
接口研究者的广泛关注。在这方面机器学习和
子计算等新工具将有助于脑机接口成为现实。而
大规模、高质量的数据集则有助于推动解码算法的
发展。清华大学研究团队发布了基于稳态视觉诱
发电位的脑机接 口的 BETA 数据集[34]。该研究具有
领域内迄今为止规模最大测试基准算法最全等
色,个体水平的脑机接口性能评估梳理了信噪
与信息传输率的关系,为群体水平的脑机接口性能
9各种模态下通信速率的比较
62
科技导报 20 21,3919 www.kjdb.org
出了 BCI quotient新指
该研 向真 的脑 -
式研究、算法开发提供了测试平大数和人
工智能趋势下的新方法、新系统研究做好了数据支
撑。 世界机器-BCI 脑控机器人大赛
的平台也极大地推动了国内脑机接口的算法水平。
2019 世界机器人大赛-BCI 脑控机器人大赛暨第
三届中国 脑机接口比赛现 场,实现了 691.55 bit/min
的理想信息传输速率,创造了历届世界机器人大会
脑控打字最高纪录10。脑机接口技术的发展
离不开领域内学者的共同努力,期待着后续更多跨
研究组、跨高校的通力合作。
3.2 双向脑机接口
在脑机交中,信息可在两个方上传播:
“从脑到机”将脑信号转换成意图运动指令“从
机到脑”与外部环交互的设备捕获的觉信
传递 脑) 领域
“从脑到机”为主在机械臂碰到物体后,受试
者只能通过视觉来了解控制的结果。近年来
调控技术的发从机脑”提供了可能调节
神经活动将是下一代脑机接口的重要组成部分,
如,神经修复运动控制提供触觉。匹兹堡大学
合作研究团队展示了通过体感皮层的皮层内微电
intracortical microstimulation
知反馈,使得具有双向脑机接口的受试者能够改善
其在由神经控制的假肢完成的功能性物体运输任
务中 的性能(图 11[35]
。受试者在视觉反馈的基础
之上结合微电刺激诱发的触觉知,能够更快地
完成任务。这些结果验证了触觉在上肢控制中的
重要性以及利用皮层内微电刺激恢复信息流以创
建双向脑机接口的效用[35]
3.3 信息安
与健备越
行。一方面,户可以方便了解自身的健康状况
息;另一方些信也面临新的安全风。乔
·华盛顿大学及合作研究团队研究了家用脑电
NeuroSky App store 中 的 156
脑机 接口应用 程序都 容易受 到近程攻 击,而且 31
个免费应用程序都容易受到至少一种远程攻击的
攻击。此外,团队提出了一种联合循环卷积神
网络的深度学习模型,能够从 NeuroSky 脑电设备
中窃取的精简特征的脑电数据推测用户活动
测精度可达 70.55%[36] 。考到脑动的度私
性和重要性,在实现脑机接口应的过程中如何
对脑活动数据进行有效安全的管理并制定相关标
准规范是当下科研界和产业界都必须深入思考的
关键一环。
4结论
20192020 年,脑机接口技术在理论分析、
件实现、算法改进应用等方面均取得阶段
性的研究进展,对推动脑机接口技术的发展起到了
10 2019 世界机器人大赛-BCI 脑控机器人大赛现场
11 双向脑机接口系统
63
科技导报 20 21,3919
www.kjdb.org
重要的作用。目前脑机接口仍主要局限于复杂的
实验室环境。对于侵入式脑机接口而言,目前仍面
临着人体排异反应及颅骨向外传输信息会减损这
两大问题非侵入式脑机接口技则朝小型化、便
携化可穿戴化及简单易用化方发展入式
的技术在应用方面更有优势。随着神经科学
接口和机器学习技术的进步,机接口领域正在
速扩展。尽管目前而言,脑机接口技术仍未达到自
然交互的沟通度和准确性,但是着各对这
项技术越来越重视,信脑机接口技术的爆发未
可期。
参考文献References
[1] Benabid A L, Costecalde T, Eliseyev A, et al. An exoskel⁃
eton controlled by an epidural wireless brain machine
interface in a tetraplegic patient: A proof-of-concept
demonstration[J]. The Lancet Neurology, 2019, 18(12):
1112-1122.
[2] He Y, Eguren D, Azorín J M, et al. Brain-machine inter
faces for controlling lower-limb powered robotic systems
[J]. Journal of Neural Engineering, 2018, 15(2): 021004.
[3] Soekadar S R, Witkowski M, Gómez C, et al. Hybrid EEG/
EOG-based brain/neural hand exoskeleton restores fully
independent daily living activities after quadriplegia[J].
Science Robotics, 2016, 1: eaag3296.
[4] Biasiucci A, Leeb R, Iturrate I, et al. Brain-actuated
functional electrical stimulation elicits lasting arm motor
recovery after stroke[J]. Nature Communications, 2018, 9
(1): 2421.
[5] Chen X, Wang Y, Nakanishi M, et al. High-speed spell⁃
ing with a noninvasive brain-computer interface[J]. Pro⁃
ceedings of the National Academy of Sciences of the Unit
ed States of America, 2015, 112(44): E6058-E6067.
[6] Anumanchipalli G K, Chartier J, Chang E F. Speech syn⁃
thesis from neural decoding of spoken sentences[J]. Na⁃
ture, 2019, 568(7753): 493-498.
[7] David A M, Leonard M K, Makin J G, et al. Real-time
decoding of question-and-answer speech dialogue using
human cortical activity[J]. Nature Communications, 2019,
10(1): 3096.
[8] Makin J G, Moses D A, Chang E F. Machine translation
of cortical activity to text with an encoder-decoder frame
work[J]. Nature Neuroscience, 2020, 23(4): 575-582.
[9] Ganzer P D, Colachis S C, Schwemmer MA, et al. Restor⁃
ing the sense of touch using a sensorimotor demultiplex⁃
ing neural interface[J]. Cell, 2020, 181(4): 763-773.
[10] Benabid A L, Costecalde T, Eliseyev A, et al. An exo⁃
skeleton controlled by an epidural wireless brain-ma⁃
chine interface in a tetraplegic patient: A proof-of-con
cept demonstration[J]. The Lancet Neurology, 2019, 18
(12): 1112-1122.
[11] Edelman B J, Meng J, Suma D, et al. Noninvasive neuro⁃
imaging enhances continuous neural tracking for robotic
device control[J]. Science Robotics, 2019, 4(31):
eaaw6844.
[12] Musk E. Neuralink: An integrated brain-machine inter
face platform with thousands of channels[J]. Journal of
Medical Internet Research, 2019, 21(10): e16194.
[13] Yamakawa T, Inoue T, Niwayama M, et al. Implantable
multi-modality probe for subdural simultaneous measure⁃
ment of electrophysiology, hemodynamics, and tempera⁃
ture distribution[J]. IEEE Transactions on Biomedical
Engineering, 2019, 66(11): 3204-3211.
[14] Hill R M, Boto E, Holmes N, et al. A tool for functional
brain imaging with lifespan compliance[J]. Nature Com
munications, 2019, 10(1): 4785.
[15] Mahmood M, Mzurikwao D, Kim Y S, et al. Fully porta
ble and wireless universal brain-machine interfaces en
abled by flexible scalp electronics and deep learning al⁃
gorithm[J]. Nature Machine Intelligence, 2019, 1: 412-
422.
[16] Kappel S L, Rank M L, Toft H O, et al. Dry-contact
electrode ear-EEG[J]. IEEE Transactions on Biomedical
Engineering, 2019, 66(1): 150-158.
[17] Lin S, Liu J, Li W, et al. A flexible, robust, and gel-free
electroencephalogram electrode for noninvasive brain-
computer interfaces[J]. Nano Letters, 2019, 19(10):
6853-6861.
[18] Degenhart A D, Bishop W E, Oby E R, et al. Stabiliza⁃
tion of a brain-computer interface via the alignment of
low-dimensional spaces of neural activity[J]. Nature Bio⁃
medical Engineering, 2020, doi: 10.1038/s41551-020-
0542-9.
[19] Zhang Y, Nam C S, Zhou G, et al. Temporally con⁃
strained sparse group spatial patterns for motor imagery
BCI[J]. IEEE Transactions on Cybernetics, 2019, 49(9):
3322-3332.
[20] Daly I, Williams D, Malik A, et al. Personalised, multi-
modal, affective state detection for hybrid brain-comput
er music interfacing[J]. IEEE Transactions on Affective
Computing, 2020, 11(1): 111-124.
[21] Jeong J H, Kwak N S, Guan C, et al. Decoding move⁃
ment-related cortical potentials based on subject-depen
dent and section-wise spectral filtering[J]. IEEE Transac⁃
tions on Neural Systems and Rehabilitation Engineering,
2020, 28(3): 687-698.
64
科技导报 20 21,3919 www.kjdb.org
[22] Zhang W, Wu D. Manifold embedded knowledge transfer
for brain-computer interfaces[J]. IEEE Transactions on
Neural Systems and Rehabilitation Engineering, 2020, 28
(5): 1117-1127.
[23] Corsi M C, Chavez M, Schwartz D, et al. Integrating
EEG and MEG signals to improve motor imagery classifi⁃
cation in brain-computer interface[J]. International Jour⁃
nal of Neural Systems, 2019, 29(1): 1850014.
[24] Willett F R, Deo D R, Avansino D T, et al. Hand knob
area of premotor cortex represents the whole body in a
compositional way[J]. Cell, 2020, 181(2): 396-409.
[25] Khalaf A, Sejdic E, Akcakaya M. A novel motor imagery
hybrid brain computer interface using EEG and function
al transcranial Doppler ultrasound[J]. Journal of Neuro⁃
science Methods, 2019, 313: 44-53.
[26] Sereshkeh A R, Yousefi R, Wong A T, et al. Online clas
sification of imagined speech using functional near-infra
red spectroscopy signals[J]. Journal of Neural Engineer⁃
ing, 2019, 16(1): 016005.
[27] Nagai H, Tanaka T. Action observation of own hand
movement enhances event-related desynchronization[J].
IEEE Transactions on Neural Systems and Rehabilita⁃
tion Engineering, 2019, 27(7): 1407-1415.
[28] Faller J, Cummings J, Saproo S, et al. Regulation of
arousal via online neurofeedback improves human perfor⁃
mance in a demanding sensory-motor task[J]. Proceed
ings of the National Academy of Sciences of the United
States of America, 2019, 116(13): 6482-6490.
[29] Rashkov G, Bobe A, Fastovets D, et al. Natural image re⁃
construction from brain waves: A novel visual BCI sys
tem with native feedback[J]. bioRxiv, 2019: 10.1101/
787101.
[30] Zhao H, Wang Y, Liu Z, et al. Individual identification
based on code-modulated visual-evoked potentials[J].
IEEE Transactions on Information Forensics and Securi
ty, 2019, 14(12): 3206-3216.
[31] Yao Z, Wang Y, Yang C, et al. An online brain-comput⁃
er interface in mobile virtual reality environments[J]. In
tegrated Computer-Aided Engineering, 2019, 26(4): 345-
360.
[32] Chen X, Hu N, Wang Y, et al. Validation of a brain-
computer interface version of the digit symbol substitu⁃
tion test in healthy subjects[J]. Computers in Biology
and Medicine, 2020, 120: 103729.
[33] Chang E F, Anumanchipalli G K. Toward a speech neu⁃
roprosthesis[J]. JAMA, 2020, 323(5): 413-414.
[34] Liu B, Huang X, Wang Y, et al. BETA: A large bench⁃
mark database toward SSVEP-BCI application[J]. Fron⁃
tiers in Neuroscience, 2020, 14: 627.
[35] Flesher S N, Downey J E, Weiss JM, et al. Restored tac⁃
tile sensation improves neuroprosthetic arm control[J].
bioRxiv, 2019: 10.1101/653428.
[36] Xiao Y, Jia Y, Cheng X, et al. I can see your brain: In⁃
vestigating home-use electroencephalography system se
curity[J]. IEEE Internet of Things Journal, 2019, 6(4):
6681-6691.
Hot topics review of brain-computer interface in 2019—2020
AbstractAbstract Brain-computer interface (BCI) is designed to provide rich and powerful command signals for assistive devices by
decoding user's intention directly from brain signals in a real-time way. Recently, both theoretic and practical aspects of BCI
technology have rapidly developed and become increasingly mature. More application scenarios of BCI technology have been
demonstrated as well. This review summarizes the important achievements and events in hardware, algorithm, paradigm, and
application in the BCI field in 20192020 and discusses its development trend.
KeywordsKeywords brain-computer interface; application system; key technology
责任编辑 刘志远
CHEN Xiaogang1, YANG Chen2, CHEN Jingjing3, GAO Xiaorong3*
1. Institute of Biomedical Engineering, Chinese Academy of Medical Sciences and Peking Union Medical College, Tianjin
300192, China
2. School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
3. Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing 100084, China
65
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
The brain-computer interface (BCI) provides an alternative means to communicate and it has sparked growing interest in the past two decades. Specifically, for Steady-State Visual Evoked Potential (SSVEP) based BCI, marked improvement has been made in the frequency recognition method and data sharing. However, the number of pubic databases is still limited in this field. Therefore, we present a BEnchmark database Towards BCI Application (BETA) in the study. The BETA database is composed of 64-channel Electroencephalogram (EEG) data of 70 subjects performing a 40-target cued-spelling task. The design and the acquisition of the BETA are in pursuit of meeting the demand from real-world applications and it can be used as a test-bed for these scenarios. We validate the database by a series of analyses and conduct the classification analysis of eleven frequency recognition methods on BETA. We recommend using the metric of wide-band signal-to-noise ratio (SNR) and BCI quotient to characterize the SSVEP at the single-trial and population levels, respectively. The BETA database can be downloaded from the following link http://bci.med.tsinghua.edu.cn/download.html.
Article
Full-text available
The instability of neural recordings can render clinical brain–computer interfaces (BCIs) uncontrollable. Here, we show that the alignment of low-dimensional neural manifolds (low-dimensional spaces that describe specific correlation patterns between neurons) can be used to stabilize neural activity, thereby maintaining BCI performance in the presence of recording instabilities. We evaluated the stabilizer with non-human primates during online cursor control via intracortical BCIs in the presence of severe and abrupt recording instabilities. The stabilized BCIs recovered proficient control under different instability conditions and across multiple days. The stabilizer does not require knowledge of user intent and can outperform supervised recalibration. It stabilized BCIs even when neural activity contained little information about the direction of cursor movement. The stabilizer may be applicable to other neural interfaces and may improve the clinical viability of BCIs. Neural activity residing in a low-dimensional space that reflects specific correlation patterns among neurons can be used to maintain the performance of brain–computer interfaces in the presence of recording instabilities.
Article
Full-text available
A decade after speech was first decoded from human brain signals, accuracy and speed remain far below that of natural speech. Here we show how to decode the electrocorticogram with high accuracy and at natural-speech rates. Taking a cue from recent advances in machine translation, we train a recurrent neural network to encode each sentence-length sequence of neural activity into an abstract representation, and then to decode this representation, word by word, into an English sentence. For each participant, data consist of several spoken repeats of a set of 30–50 sentences, along with the contemporaneous signals from ~250 electrodes distributed over peri-Sylvian cortices. Average word error rates across a held-out repeat set are as low as 3%. Finally, we show how decoding with limited data can be improved with transfer learning, by training certain layers of the network under multiple participants’ data.
Article
Full-text available
An important challenge in developing a movement-related cortical potential (MRCP)-based brain-machine interface (BMI) is an accurate decoding of the user intention for real-world environments. However, the performance remains insufficient for real-time decoding owing to the endogenous signal characteristics compared to other BMI paradigms. This study aims to enhance the MRCP decoding performance from the perspective of preprocessing techniques (i.e., spectral filtering). To the best of our knowledge, existing MRCP studies have used spectral filters with a fixed frequency bandwidth for all subjects. Hence, we propose a subject-dependent and section-wise spectral filtering (SSSF) method that considers the subjects’ individual MRCP characteristics for two different temporal sections. In this study, MRCP data were acquired under a powered exoskeleton environments in which the subjects conducted self-initiated walking. We evaluated our method using both our experimental data and a public dataset (BNCI Horizon 2020). The decoding performance using the SSSF was 0.86 (±0.09), and the performance on the public dataset was 0.73 (±0.06) across all subjects. The experimental results showed a statistically significant enhancement (p0.01) compared with the fixed frequency bands used in previous methods on both datasets. In addition, we presented successful decoding results from a pseudo-online analysis. Therefore, we demonstrated that the proposed SSSF method can involve more meaningful MRCP information than conventional methods.
Article
Full-text available
The human brain undergoes significant functional and structural changes in the first decades of life, as the foundations for human cognition are laid down. However, non-invasive imaging techniques to investigate brain function throughout neurodevelopment are limited due to growth in head-size with age and substantial head movement in young participants. Experimental designs to probe brain function are also limited by the unnatural environment typical brain imaging systems impose. However, developments in quantum technology allowed fabrication of a new generation of wearable magnetoencephalography (MEG) technology with the potential to revolutionise electrophysiological measures of brain activity. Here we demonstrate a lifespan-compliant MEG system, showing recordings of high fidelity data in toddlers, young children, teenagers and adults. We show how this system can support new types of experimental paradigm involving naturalistic learning. This work reveals a new approach to functional imaging, providing a robust platform for investigation of neurodevelopment in health and disease. Magnetoencephalography (MEG) recordings are sensitive to movement and therefore are especially challenging with young participants. Here the authors develop a wearable MEG system based on a modified bicycle helmet, which enables reliable recordings in toddlers, children, teenagers and adults.
Article
Paralyzed muscles can be reanimated following spinal cord injury (SCI) using a brain-computer interface (BCI) to enhance motor function alone. Importantly, the sense of touch is a key component of motor function. Here, we demonstrate that a human participant with a clinically complete SCI can use a BCI to simultaneously reanimate both motor function and the sense of touch, leveraging residual touch signaling from his own hand. In the primary motor cortex (M1), residual subperceptual hand touch signals are simultaneously demultiplexed from ongoing efferent motor intention, enabling intracortically controlled closed-loop sensory feedback. Using the closed-loop demultiplexing BCI almost fully restored the ability to detect object touch and significantly improved several sensorimotor functions. Afferent grip-intensity levels are also decoded from M1, enabling grip reanimation regulated by touch signaling. These results demonstrate that subperceptual neural signals can be decoded from the cortex and transformed into conscious perception, significantly augmenting function.
Article
Transfer learning makes use of data or knowledge in one problem to help solve a different, yet related, problem. It is particularly useful in brain-computer interfaces (BCIs), for coping with variations among different subjects and/or tasks. This paper considers offline unsupervised cross-subject electroencephalogram (EEG) classification, i.e., we have labeled EEG trials from one or more source subjects, but only unlabeled EEG trials from the target subject. We propose a novel manifold embedded knowledge transfer (MEKT) approach, which first aligns the covariance matrices of the EEG trials in the Riemannian manifold, extracts features in the tangent space, and then performs domain adaptation by minimizing the joint probability distribution shift between the source and the target domains, while preserving their geometric structures. MEKT can cope with one or multiple source domains, and can be computed efficiently. We also propose a domain transferability estimation (DTE) approach to identify the most beneficial source domains, in case there are a large number of source domains. Experiments on four EEG datasets from two different BCI paradigms demonstrated that MEKT outperformed several stateof- the-art transfer learning approaches, and DTE can reduce more than half of the computational cost when the number of source subjects is large, with little sacrifice of classification accuracy.
Article
Decades after the motor homunculus was first proposed, it is still unknown how different body parts are intermixed and interrelated in human motor cortical areas at single-neuron resolution. Using multi-unit recordings, we studied how face, head, arm, and leg movements are represented in the hand knob area of premotor cortex (precentral gyrus) in people with tetraplegia. Contrary to traditional expectations, we found strong representation of all movements and a partially “compositional” neural code that linked together all four limbs. The code consisted of (1) a limb-coding component representing the limb to be moved and (2) a movement-coding component where analogous movements from each limb (e.g., hand grasp and toe curl) were represented similarly. Compositional coding might facilitate skill transfer across limbs, and it provides a useful framework for thinking about how the motor system constructs movement. Finally, we leveraged these results to create a whole-body intracortical brain-computer interface that spreads targets across all limbs.
Article
Digit symbol substitution test (DSST), which is a valid and sensitive tool to assess human cognitive dysfunction, has been widely used in clinical neuropsychology. Although several versions of DSST are currently available, most of the existing DSST versions rely on examinees' intact motor function. This limits their utility in severely motor-impaired individuals. A brain-computer interface (BCI) version of DSST was implemented in this study. Steady-state visual evoked potential (SSVEP) was adopted to build the BCI. Nine symbols in the proposed SSVEP BCI-based DSST were designed with clearly different shapes for decreasing measurement errors due to misidentified symbols. To reduce practice effect, furthermore, the digit-symbol pairs of each trial were different. A two-target SSVEP BCI was designed to judge whether the digit-symbol probe in the center of the user interface matched one of the nine digit-symbol pairs above the user interface. All 12 examinees were able to perform the tasks using the proposed SSVEP BCI-based DSST with 96.17 ± 4.18% averaged accuracy, which was comparable with that of computerized DSST. Furthermore, for examinees participating in both offline and online experiment, the accuracies of the online and offline experiments were comparable, supporting that the proposed BCI-DSST was reliable for repeatedly evaluating examinees’ cognitive function over time. These results verified that the proposed SSVEP BCI-based DSST was feasible and effective for healthy subjects.
Article
Spoken communication is a basic human function. As such, loss of the ability to speak can be devastating for affected individuals. Stroke or neurodegenerative conditions, such as amyotrophic lateral sclerosis, can result in paralysis or dysfunction of vocal structures that produce speech. Current options are assistive devices that use residual movements, for example, cheek twitches or eye movements, to navigate alphabet displays to type out words.¹ While some users depend on these alternative communication approaches, these devices tend to be slow, error-prone, and laborious. A next generation of rehabilitative technologies currently being developed, called brain-computer interfaces (BCIs), directly read out brain signals to replace lost function. The application of neuroprostheses to restore speech has the potential to improve the quality of life of patients with neurological disease, but also including patients who have lost speech from vocal tract injury (eg, from cancer or cancer-related surgery).