组稀疏贝叶斯逻辑回归的P300信号通道自动选择算法

为了提高脑机接口中P300脑电信号的分类准确率和计算速度，提出一种组稀疏贝叶斯逻辑回归的P300脑电信号通道自动选择算法.该算法首先在贝叶斯框架下建立P300脑电信号的解码模型，其次提出先验的组自动相关确定(GARD)方法构建组稀疏约束下的P300脑电通道权重系数，最后通过最大似然估计来求解超参数并选出P300脑电通道最优子集，避免了大量的交叉验证过程.所提方法在BCI竞赛数据和自采集数据上进行了验证分析.实验结果表明，所提的方法能够自动选择P300脑电通道子集，提高了P300特征分类准确率.

Channel Automatic Selection Algorithm for P300 Signal with Group Sparsity Bayesian Logistic Regression

In order to improve the classification accuracy and calculation speed of P300 electroencephalogram(EEG)signals in the brain-computer interface(BCI)， a channel automatic selection algorithm of P300 EEG signal based on group sparsity Bayesian logistic regression was proposed. First， the algorithm established the decoding model of P300 EEG signals under the Bayesian framework， and then， a priori group automatic relevance determination(GARD)was proposed to determine the weight coefficients of P300 EEG channels under group sparse constraints. Finally， the maximum likelihood estimation was used to solve the hyperparameters and select the optimal subset of P300 EEG channels， avoiding a large number of cross-validation processes. The proposed method was verified on the BCI competition dataset and self-acquisition dataset. The experimental results showed that the proposed method can automatically select P300 related channels and may improve the accuracy of P300 feature classification.