一种融合注意力检测和意图识别的异步脑控方法

An asynchronous brain-computer interface method fusing attention detection and intention recognition

在机器人的异步脑控系统中,如何从脑电信号中识别使用者的控制意图状态,是当前研究中的一个难点问题。由于被试的注意力水平在控制状态和非控制状态下存在显著差异,本文提出了一种融合注意力检测和意图识别的异步脑控方法,同时检测被试的注意力水平和脑控意图并进行决策层融合,以提升整体分类性能和减少误触发。10名被试参与实验的结果表明,本文方法在融合注意力检测信息后,在数据长度为0.5 s到3 s时相比原有方法都取得了更高的分类准确率、真实阴性率和信息传输率。在数据长度1.5 s时,本文方法将平均真实阴性率从41.3%提高到了89.0%,将平均分类准确率从57.6%提高到了73.8%,将平均信息传输率从21.1 bits/min提高到了35.8 bits/min。

In an asynchronous brain-controlled robot system,one of the key issue of the current researches is to recognize the control state and the non-control state of the electroencephalogram signals.As the attentional level of the subject is significantly different between the control and non-control states,in this paper,a novel asynchronous brain-computer interface method,which performs attention detection and intention recognition in parallel and fuses their classification results in decision level,is proposed to improve the overall performance and reduce false triggers.The experimental results of 10 subjects showed that,the proposed method achieved higher classification accuracy,higher true negative rate,and higher information transfer rate than the intention detection method.When the data length is 1.5 s,the proposed method improved the true negative rate form 41.3%to 89.0%,and improved the information transfer rate from 21.1 bits/min to 35.8 bits/min.