【摘要】 近年来,稳态视觉诱发电位(steady-state visual evoked potential, SSVEP)范式脑机接口(brain-computer interface, BCI)得到了日益广泛的研究。如何选择不同的分类特征,对于提高频率识别的准确率,改善SSVEP-BCI系统至关重要。针对少目标刺激范式的SSVEP-BCI系统,提出了小波包变换(wavelet packet transform, WPT)同多变量同步指数(multivariate synchronization index, MSI)相结合的方法,对10名被试者的400组SSVEP数据进行特征提取并分类。在分类过程中,讨论了导联数量和数据长度两个参数对改进算法的影响。实验结果表明:在数据长度为1.5 s,导联7导的条件下,基于WPT-MSI的SSVEP算法的分类准确率达到98.94%,信息传输率为76.24 bit/min。明显优于典型的MSI算法和其他改进算法,具有显著提高的频率识别正确率。更多还原

【Abstract】 In recent years, a Brain-computer Interface(BCI) with a steady-state visual evoked potential(SSVEP) paradigm has been widely studied. How to choose various categorization features is very important for enhancing frequency identification accuracy and improving the performance of the SSVEP-BCI system. Aiming at the SSVEP-BCI system of the few target stimulus paradigm, a method of combining a wavelet packet transform(WPT) with a multivariate synchronization index(MSI) was proposed. Then the method was applied for feature extractions and task classifications of 400 SSVEP data sets from 10 subjects. During the classification processes, the influence of the two crucial parameters, i.e., the number of leads and the length of the data, were analyzed and discussed during the application of the improved algorithm. The results show as follows. Under the condition that the data length is 1.5 s and the lead is 7 leads, the classification accuracy of the SSVEP algorithm based on the WPT-MSI reaches 98.94% and the information transmission rate arrives 76.24 bit/min. The accuracy is significantly better than the typical MSI algorithm and other related algorithms. Thus, the accuracy of the frequency recognition of SSVEP is significantly increased by the proposed method.更多还原