中文摘要：

提出一种新的基于支持向量机（super vector machine, SVM）学习机制和数据融合理论的脑电分类算法，并设计了注意分级实验进行验证。首先，对脑电信号进行3级小波分解，由主分量分析（ principle component analysis, PCA） 方法提取其中的主特征分量；然后由支持向量机对特征分量进行分类；最后依据数据融合理论，对多导分类结果进行综合判断。结果表明，该方法具有良好的鲁棒性，对多导注意相关EEG的分类准确率可达89％左右，并高于单导最优准确率，对注意力缺陷反馈治疗、注意力机制研究等有较高的实用价值。

英文摘要：

A novel algorithm based on Support Vector Machine （SVM） and fusion rule to classify the attention-related Electroencephalogram is proposed. Wavelet Transform and Principle Component Analysis （PCA） are used to pick up main features vectors from raw EEG. Then the feature vectors are classified with SVM. After that classification result is estimated synthetically. Then the feathure vectors are classified with SVM. After classification result is estimated synthetically. The results indicate the classification accuracy of the algorithm with fusion rule reachs about 89 percent and is better than that of the best channel. At the same time, selecting suitable kernel functions can obtain better result. The experiments show the algorithm is a powerful tool for cure and research EEG patterns associated to attention tasks.