论文部分内容阅读
针对稳态视觉诱发电位(steady state visual evoked potential,SSVEP)范式下脑电信号(electroencephalograph,EEG)信噪比低,限制其识别正确率提高及脑-机接口应用等问题,根据EEG随机性、近似平稳的特点,提出了用于SSVEP特征频率提取的同步压缩短时傅里叶变换方法。该方法利用短时傅里叶变换对EEG进行时频分析,并通过同步压缩变换(synchrosqueezing transform,SST)对时频平面的能量在频率方向进行重新分配,获得频率曲线更加集中的时频表达;同时为提高EEG信噪比,提取SSVEP脑电中特征频率附近信号进行重构,并利用典型相关分析(CCA)进行分类识别,有效提高了最终识别正确率。仿真和实验结果表明,该方法极大程度地提高了信号的信噪比,具有良好的抗噪声性能和信号提取精度,且相比于传统的经验模态分解和常规滤波方法,该方法平均识别正确率最大提高分别为9.98%和4.38%,平均信息传输率最大提高分别为7.57 bit/min和2.69 bit/min,有效增强了SSVEP范式下脑-机接口的工作性能和实际应用。
Aiming at the problems such as low signal-to-noise ratio of electroencephalograph (EEG) in steady state visual evoked potential (SSVEP) paradigm, limited recognition rate and application of brain-computer interface, according to EEG randomness, Approximate smooth feature, a synchronous compression short-time Fourier transform method for feature frequency extraction of SSVEP is proposed. In the method, the time-frequency analysis of EEG is carried out by using short-time Fourier transform, and the energy of the time-frequency plane is redistributed in the frequency direction by the synchrosqueezing transform (SST) to obtain a more concentrated time-frequency expression of the frequency curve. At the same time, in order to improve the signal-to-noise ratio of EEG, the signals near the characteristic frequency in SSVEP EEG were extracted and reconstructed, and the canonical correlation analysis (CCA) was used to classify and identify them, which effectively improved the accuracy of final recognition. Simulation and experimental results show that this method can greatly improve the signal-to-noise ratio of the signal, has good anti-noise performance and signal extraction accuracy, and compared with the traditional empirical mode decomposition and conventional filtering methods, The maximum correct rate was 9.98% and 4.38%, respectively. The maximum average information transfer rate was 7.57 bit / min and 2.69 bit / min respectively, which effectively enhanced the performance and practical application of brain-computer interface in SSVEP paradigm.