摘要
基于多分类运动想象的在线BCI(brain computer interface,脑机接口)中,如何实时处理高速EEG(electroencephalogram,脑电)数据流是实现在线意识识别的难点,其关键是高速计算和复杂情况下的预测问题。以线程并发作为解决高速计算问题的切入点,首先将EEG信号分析任务分解为多个线程子任务,并通过缓冲区管理策略解决线程并发带来的协同问题,针对高速EEG数据流的复杂变化问题,采用自适应单向模糊推理的方法预测数据流伸缩变化,并针对线程并发造成的中间结果的错序问题,设计信号量互斥与同步方法对中间数据块进行顺序重组。针对多名受试者的大量实验显示,单次Trial平均延迟时间明显减少。因此,线程并发和模糊推理能够解决在线BCI系统的高速计算和预测问题,从而提高信息传输率。
About online BCI based on multi-class motor imagery, how to handle high-speed EEG data streams is a difficulty for the realizing of online awareness recognition, and the key is high-speed computing and prediction under complicated condi- tions. This paper took thread concurrency as the entry point of high-speed computing, firstly, it decomposed the task of EEG signal analysis into more thread subtasks, and solved the coordination problem brought by thread concurrency with buffer man- agement policies; then, for the complicated change of high-speed EEG data streams, it adopted adaptive one-sided fuzzy infer- enee to predict the telescopic change of data streams; lastly, against the disorders of intermediate result due to thread concur- rency, it designed a method of mutual exclusion and synchronization with semaphore to recombine the intermediate data blocks orderly. Numerous experiments with multiple subjects show that the average delay time of a single Trial decreases obviously. Therefore, thread concurrency and fuzzy inference can solve the problem of high-speed computing and prediction in online BCI, and improve the information transmission rates.
出处
《计算机应用研究》
CSCD
北大核心
2015年第3期794-799,共6页
Application Research of Computers
基金
国家自然科学基金资助项目(60841004
60971110
61172152)
郑州市科技攻关(112PPTGY219-8)
河南省青年骨干教师计划资助项目(2012GGJS-005)
关键词
在线BCI
高速EEG数据流
并发
自适应单向模糊推理
生产—消费协同
online BCI
high-speed EEG data streams
concurrency
adaptive one-sided fuzzy inference
coordination ofproduction -consumption
