采用最小二乘转换的在线SSVEP字符输入系统设计与实现  

Design and implementation of online SSVEP character inputsystem using least square transformation

作　　者：李振华 刘柯 邓欣[1] LI Zhenhua;LIU Ke;DENG Xin(School of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China)

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2022年第5期859-868,共10页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金(61703065);重庆市基础研究与前沿探索项目(cstc2018jcyjAX0151);重庆市自然科学基金(cstc2020jcyj-msxmX0284);重庆市教委科技项目青年项目(KJQN202000625)。

摘　　要：针对SSVEP-BCI系统信息传输率高、鲁棒性强,当前基于有训练分类算法的SSVEP-BCI系统需要较长时间采集训练数据,而基于无训练算法的系统难以满足实时性要求的问题,提出只需少量训练数据的高效在线字符输入系统,实现了快速准确的字符输入。该方法利用最小二乘转换技术进行跨被试迁移学习,并使用FoMSFA和多频率学习技术进行频率识别。使用者仅需进行2组训练数据采集,即可在1.96 s内实现单字符快速准确地在线输入。对15名受试者执行有提示字符输入,平均准确率和信息传输率分别为79.3%和161.9 bit/min;10名受试者执行无提示字符输入,平均准确率和信息传输率分别为80.0%和163.5 bit/min的实验。迁移学习技术和高效SSVEP识别算法的结合,为在线SSVEP-BCI系统的发展提供了新思路。The SSVEP-BCI system has high information transmission rate and strong robustness.However,the SSVEP-BCI system based on frequency recognition algorithms required a long time to collect training EEG data,and the system based on training-free classification algorithms is difficult to meet the real-time requirements.This paper proposes an efficient online character input system that requires only a small amount of training data,realizing fast and accurate character input.Specifically,the proposed method employs the least square transformation technology for cross-subject transfer learning,and uses FoMSFA and multi-frequency learning technique for frequency recognition.Subjects can input a single character accurately within 1.96s with only 2 sessions of training data collection.15 subjects were given character input with cues,and the average accuracy rate and information transmission rate were 79.3%and 161.9 bit/min,respectively;10 subjects were given character input without cues,achieving the average accuracy rate with 80.0%and information transmission rate 163.5 bit/min,respectively.The combination of transfer learning technology and efficient recognition algorithm provides a new idea for the development of the online SSVEP-BCI system.

关 键 词：脑机接口 稳态视觉诱发电位 在线字符输入系统 迁移学习 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] R318[自动化与计算机技术—控制科学与工程]