论文部分内容阅读
目的研究一种对多任务脑电信号进行快速识别的自适应分类算法,提高脑-机接口(brain-computer interface,BCI)系统的实用性。方法将孪生支持向量机(twin support vector machine,TSVM)作为初始分类模型,通过后验概率输出建模方法求得新样本属于各个类别的概率,并将该样本归为概率最大的类别,然后引用增量学习方法将满足一定条件的新样本加入到训练集中来更新分类模型,以最新的分类模型对新增样本进行识别。结果对2008年BCI竞赛数据集Dataset1和Dataset2a进行分类,与传统SVM和现有TSVM等方法相比,该方法降低了分类耗时,能更好地识别出多数受试者的脑电信号。结论本文算法能提高分类器的自适应性和分类速度,为BCI系统提供了一种有效的在线识别方法。
Objective To study an adaptive classification algorithm for rapid recognition of multitask EEG signals to improve the practicality of a brain-computer interface (BCI) system. Methods The twin support vector machine (TSVM) is used as the initial classification model, and the posterior probability output modeling method is used to determine the probability that the new sample belongs to each category, and the sample is classified as the category with the highest probability. The incremental learning method adds new samples satisfying certain conditions to the training set to update the classification model and identifies the new samples with the latest classification model. Results The data of Dataset1 and Dataset2a in 2008 were classified. Compared with the traditional SVM and the existing TSVM, this method reduced the classification time and better recognized the EEG signals of most subjects. Conclusion This algorithm can improve the self-adaptability and classification speed of classifier, and provide an effective method for online identification of BCI system.