Investigating novel approaches to the real time classification of non-invasive brain recordings for use in assistive technologies

脑机接口（bci）是辅助技术（AT）的一种形式，可以用于严重运动或感觉残疾的人。然而，对于目前的非侵入性脑机接口，用户需要在期望的任务上投入大量的注意力，这有效地消除了他们进行多任务和参与任何其他并发交互行为的机会。对bci控制高度关注的需求限制了此类设备支持用户独立性的机会。使用直观且无需用户培训的脑机接口非常适合用于家庭环境控制、社交互动和参与、智能轮椅导航和指导机器人个人助理。在这个项目中，我们将进一步推进我们对非侵入性，实时BCI控制器的研究，该控制器利用头皮记录的脑电图（EEG）信号，可以识别预期运动动作的特征，并且适合集成到智能AT控制器中，具有快速和强大的事件分类潜力。对BCI事件指令的用户发起的大脑特征进行实时分类将需要对信号本身进行研究，最适合实时分类的信号处理算法，实现这一目标的计算机体系结构，以及开发半自动AT，使用户不那么关注BCI/AT本身。因此，该项目的目标是开发和实施共享（协作）控制策略，旨在减少脑机接口连接at用户的认知负荷和注意力。该项目将开发人类神经生理记录的实时信号捕获、处理和分类的研究技能。将人工智能应用到先进辅助设备模拟器（电动轮椅）的现有模型系统中，以便允许对控制器进行虚拟测试，其中BCI元素和设备具有基于1)对意图事件（BCI）的鲁棒和快速检测以及2)将这些命令转换为面向目标的高级请求（导航）到半自主AT设备的共享控制。现有的仿真器将为这项工作的共享控制策略的开发提供基础。最终目标是在实际的AT上实现这些策略，并在实际情况下测试其性能。

Brain Computer Interfaces (BCIs) are a form of Assistive Technology (AT) that can to be used by people with severe motor or sensory disability. However, with current non-invasive BCIs there is a need for the user to devote a significant level of attentive effort on the desired task which effectively abolishes the opportunity for them to multitask and engage in any other concurrent interactive behaviour. The need for high levels of attention to be directed the control of BCIs limits the opportunity for such devices to support user independence.BCIs that are intuitive to use and require little user training are highly desirable for use in home environmental control, social interaction and engagement, navigating smart wheelchairs and instructing robot personal assistants.In this project we will further advance our research into non-invasive, real time BCI controllers that exploit scalp recorded electroencephalographic (EEG) signals that can recognise signatures for intended motor actions and that are suited for integration into smart AT controllers with the potential for fast and robust event classification. Real time classification of user initiated brain signatures for event instructions to a BCI will require research on the signals themselves, the signal processing algorithms that are best suited to classification in real time, the computer architectures to achieve this and development of semi-autonomous AT that will enable the user to be less attentive to the BCI/AT itself. Accordingly, the objective of this project is to develop and implement shared (collaborative) control strategies with the aim to reducing cognitive load and attention of users of BCI linked ATs.The project will develop research skills in real time signal capture, processing and classification of human neurophysiological recordings. The implementation of artificial intelligence into an existing model system of an advanced assistive device simulator (electric wheelchair) in order to allow virtual testing of controllers in which the BCI element and the device have shared control based on 1) the robust and rapid detection of intention events (BCI) and 2) the translation of these commands into goal orientated high, level requests (navigation) to the semi-autonomous AT device. The existing simulator will provide the foundation for development of the shared control strategies for this work.The ultimate goal will be to implement these strategies onto an actual AT and test its performance in real world situations.