【摘要】 在多模态脑机接口系统(Brain-Computer Interface, BCI)的研究中，有特征级、决策级以及数据级三种融合策略。数据级融合在特征提取之前将多种模态数据融合，能够提升系统计算效率。但是，由于不同模态信号的采样率和幅值尺度等存在差异，给数据级融合的实现带来挑战。提出了一种基于小波变换的脑电(Electroencephalogram, EEG)和功能性近红外光谱(Functional Near-Infrared Spectroscopy, fNIRS)多模态数据融合方法。首先对两种信号进行小波分解，再将两组小波系数通过基于Fisher值的融合规则生成新的小波系数，然后通过小波重构得到融合信号，最后提取融合信号的共空间模式特征，利用线性判别分析进行分类。在对心理算数任务数据的分类实验中，获得88.1%的分类精度，表明了所提出方法的有效性和鲁棒性。更多还原

【Abstract】 In hybrid brain-computer interface(BCI)studies, there are three fusion strategies, i.e.,feature-level fusion, decision-level fusion and data-level fusion. Multiple modal data are fused before feature extraction in data-level fusion, which can improve the computational efficiency of the system. However, due to the difference of sampling rate and amplitude scale among different signals, the effective implementation of data-level fusion has brought challenges. A method for multimodal data fusion of Electroencephalogram(EEG)and Functional near-infrared spectroscopy(fNIRS)based on wavelet transform is proposed. Firstly, the two types of signals are decomposed into two sets of wavelet coefficients. Then, the coefficients are reconstructed into the fusion signal according to Fisher value. Finally, the common space pattern features of the fused signals are extracted and classified by using linear discriminant analysis. In the classification experiment of mental arithmetic task data, the classification accuracy of 88.1% is obtained, which shows the effectiveness and robustness of the proposed method.更多还原