首页>
外国专利>
METHOD OF CALIBRATION OF A DIRECT NEURONAL INTERFACE BY PENALIZED MULTIVOIE REGRESSION
METHOD OF CALIBRATION OF A DIRECT NEURONAL INTERFACE BY PENALIZED MULTIVOIE REGRESSION
展开▼
机译:精确的多元语音回归校正直接神经元界面的方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
The invention relates to a method for calibrating a direct neural interface (BCI). The BCI interface receives electro-physiological signals and provides control signals describing a path to a computer or machine. The electrophysiological signals are represented by an input tensor and the path by an output tensor, the interface performing an estimate of the output tensor from the input tensor based on a linear predictive model. The input tensor is extended according to the mode of the observation instants to take into account the derivative of the components of the input tensor and / or a polynomial interpolation of these components. The parameters of the linear predictive model are computed during a learning phase from a partial least squares (NPLS) multivariate regression between the output tensor and the thus extended input tensor.
展开▼