Long-term Robustness of Brain-Machine Interfaces

这项研究项目涉及脑机接口(BMI)，这是一种使用对受试者大脑的测量来提供他们与计算机交互的机制的设备。当前BMI方法可行性的主要障碍之一在于操作环境的变化：与所需行为相关的神经模式在新的记录条件下可以不同地表现，甚至可能由于所涉及的神经元的重新配置而随着时间的推移而发生根本变化。这些问题普遍存在于各种BMI设计中，尽管这些设计的工作原理依赖于不同类型的测量或信号处理。这个项目试图通过各种方法来解决这个障碍并提高BMI的长期健壮性：通过开发新的数字信号处理技术来更好地适应可变的操作环境，通过提高我们对神经可变性的基本机制的理解以使自适应解码器能够连续地校准自己，以及通过调查大脑本身需要内部校正这种可变性的反馈系统的性质。该项目本质上是跨学科的，涉及神经科学和生物信息学、控制工程、数字信号处理和机器学习等领域。

This research project deals with Brain-Machine Interfaces (BMIs), devices that use measurements of a subject's brain to provide a mechanism for them to interact with a computer. One of the main obstacles to the viability of current BMI approaches lies in the changing nature of the operating environment: the neural patterns associated with a desired behaviour can manifest differently under new recording conditions, or may even fundamentally vary over time due to reconfiguration of the neurons involved. These issues are prevalent across various BMI designs, even though the operating principles of those designs rely on different kinds of measurements or signal processing. This project seeks to tackle this obstacle and improve the long-term robustness of BMIs through a variety of approaches: by developing novel digital signal processing techniques that better adapt to the variable operating environment, by improving our understanding of the underlying mechanics of neural variability to enable adaptive decoders to continuously calibrate themselves, and by investigating the nature of the feedback systems the brain itself requires to internally correct for this variability.The project is inherently cross-disciplinary, and touches upon areas of Neuroscience and Biological Informatics, Control Engineering, Digital Signal Processing, and Machine Learning.