METHOD FOR CALIBRATING A DIRECT NEURAL INTERFACE BY PENALISED MULTI-CHANNEL REGRESSION

摘要

the invention relates to a calibration method of a direct brain computer interface (bci). if the interface receives signals of electrophysiological and provides control signals to a computer of a trajectory or a machine.the electro physiological signals are represented by a tensor of entry and exit path by a tensor, the interface with an estimate of the tensor, tensor exit from the entry on the basis of a linear prediction model.the tensor of entry is extended according to the method of moments of observation in order to take into account the derivative of tensor components of the entry and / or polynomial interpolation of these components.the parameters of the linear prediction model are determined in a learning phase using a multivariate regression, partial least squares (npls) between the tensor and tensor of entry and exit.