From adaptive signal processing to brain-inspired cognitive systems

该计划的长期研究目标（愿景）是通过引入统一而强大的信息处理框架来显著推进信号处理的研究。为了实现这一愿景，该程序引入了一个基于模型的框架，该框架由本地、处理单元、推理/适应引擎和控制单元组成。*这种受大脑启发的认知框架采用测量模块来最大限度地从环境中获取信息，使用基于记忆的注意机制来处理信息，部署推理/决策引擎来识别不确定环境中的智能选择，并利用反馈控制以高效和经济的方式与环境进行交互。*******该研究计划将证明所提出的框架将信号/信息处理问题分解为一组大大降低复杂性的处理任务，这些任务更容易设计和实现。因此，它构成了研究无监督学习、最优估计、决策和控制等挑战性问题的自然环境。此外，它将表明该框架既可以应用于研究良好的线性问题，也可以应用于以非线性、非高斯、非平稳条件为特征的复杂学习问题。*此外，除了具有重要的理论意义外，该计划的结果将有助于推进跨学科领域的最新技术，如神经科学数据处理，并推进促进福祉和提高生活质量的技术解决方案。为此，将使用基于脑电图的脑机交互（BMI）领域的两个开放研究问题（挑战）来演示该框架的实用性。也就是说，该框架将用于：(i)调和和统一认知电生理学中使用的空间过滤器，以及（ii）展示如何将这些空间过滤器用于康复系统的所谓零训练BMI接口。*******该计划的智力优点是开发了一个无所不包的认知信息处理框架。它将产生新的结果，对方法的进步，对大脑启发信息处理的理论理解，以及在实际应用中有用的解决方案具有重要意义。认知信号处理领域的创新将远远超出工程界的应用范围，对生活质量和医疗保健产生深远影响。******最后，该计划将为参与的学生提供一个培训和发展经验的平台，并将为进一步研究认知系统提供重要的灵感。它将有助于围绕认知信息处理的特定兴趣，形成新的研究贡献者和开发人员社区

The long term research objective (vision) of this program is to significantly advance research in signal processing by introducing a unifying and powerful information processing framework. To achieve this vision the program introduces a model-based framework which is comprised of local, processing units, an inference/adaptation engine, and a control unit.*This brain-inspired, cognitive framework employs a measurement module to maximize information gain from the environment, uses memory-based attentional mechanisms to process information, deploys a reasoning/decision making engine to identify intelligent choices in an uncertain environment, and utilizes feedback control to interact with the environment in an efficient and cost effective manner.*******The research program will demonstrate that the proposed framework decomposes the signal/information processing problem into a set of considerably reduced complexity processing tasks which are far easier to design and implement. Thus, it constitutes a natural setting for studying challenging problems in unsupervised learning, optimal estimation, decision and control. Moreover it will be shown that the framework can be applied to both well-studied linear problems as well as complex learning problems characterized by non-linear, non-Gaussian, non-stationary conditions.*In addition, to having significant theoretical implications, the results of the program will contribute to the advancement of the state-of-the-art in interdisciplinary areas such as neuroscience data processing, and advance technological solutions which promote well-being and enhance quality of life. To that end, two open research problems (challenges) from the field of EEG-based brain-machine interaction (BMI) will be used to demonstrate the framework's utility. Namely, the framework will be used to: (i) reconcile and unify spatial filters used in cognitive electrophysiology, and (ii) show-case how these spatial filters can be used in the so-called zero-training BMI interfaces for rehabilitation systems.*******The program's intellectual merit is the development of an all-encompassing cognitive information processing framework. It will produce novel results of importance to the advancement of methodology, theoretical understanding of brain-inspired information processing, and useful solutions in its practical applications. Innovations in the areas of cognitive signal processing will have applications far beyond the engineering community with profound impact on quality of life and healthcare. ******Lastly, the program will provide a platform for training and development experience for the students involved, and it will be a major inspiration for further studies in cognitive systems. It will help spring new communities of research contributors, and developers, around specific interests in cognitive information processing.***