Proximal SVM在脑功能分类中的应用研究

Classification and recognition of brain function based on proximal support vector machine

为了研究PSVM分类器用于脑功能识别的有效性与优越性,对脑功能识别做出了深入的研究和分析。采用三名受试者在睁眼和闭眼状态下的脑电实测数据,从不同角度深入分析和比较了PSVM分类器与标准SVM分类器的性能,主要衡量指标为识别率和训练时间。结果PSVM分类器优于标准SVM分类器之处在于,在保证识别率的同时,计算速度有了显著地提高。并且随着样本维数的增加,PSVM分类器的计算速度并没有下降。PSVM用于脑电信号功能识别是高效率的,这对今后的有实时要求的脑功能分类识别问题具有重要意义。

This paper discusses the validity and advantage of using PSVM classifier in brain function recognition,and provides a thorough study and analysis on it.To adopt raw EEG data from three volunteers in eyes open and close states,analyzes and compares the performance of PSVM classifier and standard SVM classifier in different points of view,the criterion are training time and correct recognition ratio.PSVM is more efficient than SVM in the computing speed without reducing the correct recognition ratio.Even more,along with the increasing of the sample dimension,the training time is not on the rise.PSVM is efficient and with predominant performance in EEG signals processing,which is of significance in real-time brain function pattern recognition.