基于多尺度注意力时序编码网络的语音诱发脑电解码

姚梓豪,贾海蓉,李雅荣,陈桂军

Speech-evoked EEG decoding based on Multi-scale Attention Temporal Encoding Network

Zihao YAO,Hairong JIA,Yarong LI,Guijun CHEN

表 4 MATE-Net 与现有主流解码方法的性能对比

Tab.4 Performance comparison between MATE-Net and existing decoding methods

模型名称 Acctrain/% ρ r LDA 56.61 0.645 0.661 Logistic Regression 55.39 0.640 0.658 Decision Tree 61.25 0.726 0.765 Shallowconv[27] 55.15 0.493 0.459 deepconv[27] 57.49 0.648 0.703 EEGNet[27] 58.31 0.541 0.570 EEGItnet[28] 54.66 0.532 0.542 EEGTrans[29] 59.46 0.683 0.713 EEGTcNet[30] 61.80 0.712 0.738 EEGformer[31] 64.39 0.813 0.872 平均值 74.30 0.884 0.942