A Comparison of Univariate, Multivariate, Bilinear Autoregressive, and Bandpower Features for Brain-Computer Interfaces
A Comparison of Univariate, Multivariate, Bilinear Autoregressive, and Bandpower Features for Brain-Computer Interfaces
复制标题
脑机接口的单变量、多变量、双线性自回归和带功率特征的比较
DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
C. Neuper
中科院分区：
文献类型：
--
作者：
C. Brunner;M. Billinger;C. Neuper
简介 脑机接口 (BCI) 中的信号处理工具链由影响系统整体性能的多个组件组成。一般来说，原始脑电图 (EEG) 信号首先使用时间或空间滤波器进行预处理。接下来，在最终分类阶段之前提取特征。之后，可以导出控制信号。此离线分析的目的是评估自回归 (AR) 模型派生的不同特征类型对 BCI 整体性能（通过分类准确性衡量）的影响。我们假设多元 AR (MVAR) 和双线性 AR (BLAR) 参数可以产生更高的分类精度，因为它们包含有关基础信号的更多信息。与单变量 AR (UVAR) 参数相比，MVAR 参数还描述单通道之间的关系，而 BLAR 参数可以对某些非线性信号进行建模。方法 我们使用 2008 年 BCI 竞赛的数据集 2A [1]。九名受试者在不同日期参加了两次会议。基于提示的范例涉及四种不同的运动想象任务（左手、右手、脚、舌头）。一个会话包括 6 次运行，每次 48 次试验（四个类别各 12 次）。我们使用来自三个双极电极 C3、Cz 和 C4 的信号。从这三个 EEG 通道中，我们提取了不同的特征：(1) 每个通道的单变量 AR (UVAR) 参数，
Introduction The signal processing toolchain in a brain-computer interface (BCI) consists of several components that influence the overall performance of the system. In general, raw electroencephalographic (EEG) signals are first preprocessed with temporal or spatial filters. Next, features are extracted before the final classification stage. After that, a control signal can be derived. The aim of this offline analysis was to assess the influence of different feature types derived from autoregressive (AR) models on the overall performance of a BCI (measured by classification accuracy). We hypothesized that multivariate AR (MVAR) and bilinear AR (BLAR) parameters could yield higher classification accuracies because they contain more information about the underlying signals. In contrast to univariate AR (UVAR) parameters, MVAR parameters also describe the relationships between single channels, and BLAR parameters can model certain nonlinear signals. Methods We used data set 2A from the BCI Competition 2008 [1]. Nine subjects took part in two sessions on different days. The cue-based paradigm involved four different motor imagery tasks (left hand, right hand, foot, tongue). A session consisted of 6 runs with 48 trials each (12 for each of the four classes). We used signals from three bipolar electrodes C3, Cz, and C4. From those three EEG channels, we extracted different features: (1) univariate AR (UVAR) parameters for each channel,