AABAC: Adaptive Asynchronous Brain-Actuated Control

该项目旨在开发一种新型的自适应和异步脑机接口(BCI)系统，用于智能系统和机器人的脑驱动控制。最近的科学技术进步揭示了将人脑与智能机器融合起来执行最先进的自主机器无法承担的具有挑战性的任务的可能性。脑-机接口是实现这一目标的关键技术之一。BCI系统检测和分析脑波，例如脑电(EEG)信号，以了解用户的精神状态，然后将精神状态翻译成用于与计算机、机器人和其他系统通信和控制的命令。目前几乎所有基于脑电信号的高精度脑-机接口系统都使用同步协议，只能识别两种心理状态。它们的缺点是信息传输率低，用户界面不自然，这严重限制了脑机接口系统的实际应用。在此开发的方法将通过对智能轮椅和机械臂的实时脑驱动控制进行广泛的实验来评估。

This proposed project aims to develop a novel adaptive and asynchronous brain-computer interface (BCI) system for brain-actuated control of intelligent systems and robots. Recent advances in science and technology have shed light on the possibility of fusing human's brain with intelligent machines to carry out challenging tasks that the state of the art autonomous machines cannot undertake. BCI is one of the key technologies to make this possible. A BCI system detects and analyses brain waves, e.g., electroencephalography (EEG) signals, in order to understand a user's mental states, and then translates the mental states into commands for communicating with and controlling computers, robots, and other systems. Almost all the current EEG-based BCI systems of high accuracy use synchronous protocols and recognise two mental states only. Their disadvantages include low information transfer rate and unnatural user interface, which impose severe limitations on BCI systems for real-world applications. Based on our previous research in BCI and related areas, we believe that it is now very timely to develop adaptive and asynchronous BCI systems that not only have the advantages of using asynchronous protocols, such as high information transfer rate and natural operation mode, but also benefit from adaptive learning so as to improve the system's accuracy and robustness. Apart from adaptive learning, in order to achieve high accuracy and robustness, this proposed programme will investigate novel effective indicators for onset detection and optimal timing schemes for asynchronous mental state classification, discover or invent new feature spaces on which it would be easier to classify EEG patterns, and develop new methods for increasing the number of control commands mapped from a limited number of mental states. The methods developed hereby will be assessed through extensive experimentation with real-time brain-actuated control of an intelligent wheelchair and a robotic arm.