文章摘要

文章信息/Info

Title:: Preprocessing filter design and experimental evaluation of steady-state visual evoked potential signals

Author(s):: LI Weilong; ZHANG Ruizhi; GAO Guoya; LIU Peijun; GUI Zhiguo; SHANG Yu; （National Key Laboratory of Dynamic Testing Technology, North University of China, Taiyuan 030051, China）

Keywords:: Steady-state visual evoked potentials; Brain-computer interface; Preprocessing filter design; Grid search method; ANOVA test

摘要:: 为实现低噪声、高质量的脑电信号传输,增强稳态视觉诱发电位(steady-state visual evoked potentials, SSVEP)脑机接口系统的应用能力，本研究设计了一种高效的SSVEP信号预处理滤波方法，用于去除伪迹与信号漂移。首先，本研究基于前期研究的含噪数据，进行了离线分析验证，并使用网格搜索方法，优化各滤波环节中的参数设定；然后，设计了38个刺激目标的拼写器，并纳入20名健康受试者进行在线分析。结果表明，在2、3 s时间窗下，经预处理滤波后的信号分类准确率分别提升了8.88%与6.66%，信息传输速率分别提升了19.59%与10.5%。此外，将受试者分为初试组与经验组，并进行配对单因素方差检验，证实两组准确率具有显著差异(P_{2s_acc}=0.02, P_{3s_acc}=0.028)。本研究的预处理滤波方法可高效去除SSVEP信号在时域与频域中的伪迹与噪声，提高分类识别的准确率，能够对其他范式脑机接口的信号预处理工作提供参考。

Abstract:: In order to achieve low-noise and high-quality EEG signal transmission, and enhance the application capability of steady-state visual evoked potentials (SSVEP) brain-computer interface system, we designed an SSVEP signal preprocessing filter for removing artifacts and signal drift. Firstly, based on the noisy data of preliminary study, we conducted offline analysis and verification, and used the grid search method to optimize the parameter setting in each filtering step. Then, spellers with 38 stimulus targets were designed and 20 healthy subjects were included for online analysis. The results showed that under the time window of 2、3 s, the preprocessed filtered signals were improved in classification accuracy by 8.88% and 6.66%, and the information transfer rate was improved by 19.59% and 10.5%, respectively. Moreover, the subjects were divided into the naive group and the experienced group, and the paired one-way ANOVA test confirmed that there was a significant difference in the accuracy between the two groups. The preprocessing filtering method can efficiently remove the artifacts and noise of SSVEP signals in both time and frequency domains, improve the accuracy of classification recognition, and can provide a reference for signal preprocessing in other paradigms of brain-computer interfaces.