摘要
脑机接口研究涉及神经科学、信号检测、信号处理、模式识别等多学科的交叉领域,因其不断增长的潜在需求、广泛的应用前景和深入的理论研究意义,正成为一个新的研究热点。面向脑机接口大赛运动想象脑电数据的基础上,深入分析了小波变换算法的应用,对小波Mallat分解算法等通过软件方式进行了探索,在此基础上进行了专用电路实现。具体而言,在软件上实现了主成分分析(principal component analysis,PCA)算法及线性判别式分析(linear discriminant analysis,LDA)算法,获取训练样本均值矩阵、分类效果最好状态下的PCA维数及其相应的PCA投影矩阵,并确定分类时的LDA投影方向及判别阈值。在硬件上,将上述相应参数固化到加速芯片的专用存储中,以现场可编程门控为载体,实现了Mallat分解算法,PCA和LDA算法并系统验证了设计正确性。三个算法专用电路模块构成了一个脑机接口系统芯片,实现运动想象脑电EEG信号分类功能,分类准确率达到96.4%。
Brain-computer interface(BCI)research involves neuroscience,signal detection,signal processing,pattern recognition and other multidisciplinary fields,which is becoming a new research hotspot because of its growing potential demand,broad application possibility and great theoretical significance.Based on the motion imagery electroencephalogram(EEG)data of the BCI competition,the application of wavelet transform algorithm is explored deeply,and the wavelet Mallat decomposition algorithm is investigated through software,and then a special circuit is implemented upon this investigation.Specifically,the principal component analysis(PCA)algorithm and the linear discriminant analysis(LDA)algorithm are implemented in software to obtain the mean matrix of the training sample,the PCA dimension with the best classification effect and the corresponding PCA projection matrix,and determine the LDA projection direction and discriminant threshold during classification.Accordingly with the above mentioned corresponding parameters is downloaded into the dedicated chip memory in hardware,the Mallat decomposition algorithm,both PCA and LDA algorithm are implemented on field programmable gate array and the design correctness is verified by the designed system.Three algorithm oriented circuit modules together form a BCI SoC chip,which realizes the motion imagery EEG signal classification function and achieves a classification accuracy of 96.4%.
作者
江先阳
容源
JIANG Xianyang;RONG Yuan(School of Physics and Technology,Wuhan University,Wuhan 430072,China;National Physics Experimental Teaching Demonstration Center,Wuhan University,Wuhan 430072,China)
出处
《微纳电子与智能制造》
2022年第3期99-108,共10页
Micro/nano Electronics and Intelligent Manufacturing
基金
国家自然科学基金(61072135,81971702)
中央高校基本科研业务费专项(2042017gf0075,2042019gf00720)
湖北省自然科学基金(2017CFB721)
第四批“武大通识3.0”课程(2021-ybts-10)项目资助
关键词
脑机接口
小波变换
PCA算法
LDA算法
脑电信号分类
brain-computer interface
wavelet transformation
PCA algorithm
LDA algorithm
EEG signal classification
