摘要
提出了一种基于独立分量分析的支持向量机集成学习算法,用于脑机接口中P300字符识别。首先由P300信号分解出独立分量,基于Bagging算法送入支持向量机基分类器进行集成学习,通过平均的方法获得对应类别概率进行分类决策。数据来源于P300字符拼写实验,不同导联和不同序列的分类结果表明,该分类算法学习效率和分类精度高,全导联平均分类精度为96.6%,在序列数较少的情况下,平均分类精度也达到91.5%。较其他算法,识别性能好,是脑机接口的实用分类算法。
Support vector machine ensemble learning algorithm based on independent component analysis is presented to apply for P300 speller in brain computer interface. Firstly, P300 signal is decomposed into independent components, an ensemble of classifiers approach is based on Bagging algorithm, each classifier is composed of SVM trained on a small part of available data. Secondly, it obtains corresponding classification probability and implements classification discriminant with classifier output averaging. The data are from P300 speller paradigm, classification results of different channels and sequences indicate that this algorithm yields better performance for efficiency and precision, classification performance for 64 channels is obtained for mean 96.6%, classification performance is also 91.5% for five sequences. Compared with other algorithms, this practical approach has achieved better performance for BCI classification.
出处
《电子测量与仪器学报》
CSCD
2011年第11期940-945,共6页
Journal of Electronic Measurement and Instrumentation
基金
上海高校选拔培养优秀青年教师科研专项基金(5108508001)
关键词
独立分量分析
支持向量机集成
BAGGING
脑机接口
brain computer interface
support vector machine ensemble
Bagging
brain computer interface
