神经元群体解码方法及其在脑-机接口中的应用

Neural population decoding and its applications in brain-computer Interfaces

基于脑神经元放电信号的脑-机接口(brain-computer interface,BCI)系统近年来有了越来越深入的研究,它使BCI在皮层运动控制等方面更加精确、迅速。从神经工程角度,此类BCI的实现不仅依赖于多电极神经记录硬件技术的发展,还依赖于其软件技术的核心神经元群体解码方法。本文综述了目前神经元群体解码方法中已成功运用于BCI研究的四类主要算法:群矢量算法、最佳线性估计、卡尔曼滤波法、贝叶斯方法。

Brain-computer interfaces （BCIs） based on the neuronal firing activity have been a more in-depth study in recent years than ever before, for its high speed and accuracy in BCI applications as used in cortical movement control and so on. On the aspect of neural engineering, this type of BCI not only depend on the improvement of multi-electrode recording technique as its essential hardware, but also employs neural population decoding methods as a crucial process to analyze the data recorded by multielectrodes, the kernel part of its software. Four of major algorithms currently used in neural population decoding, including population vector algorithm （PVA）, optimal linear estimator （OLE）, Kalman filter, and Bayesian decoding, are reviewed and compared in this article. All of these methods have been successfully applied in the BCI studies.