Figure 1
Description of the coherence-based spectro-spatial filter. The diagram represents the recordings of a trial while the patient performs a movement with his/her thumb. (A) Brain features are used as input to a set of linear filters trained for each electrode. The filters represent the transfer function of a linear system that maps the brain features to the movement of the finger. The outputs of all filters are combined using linear regression to produce the final prediction of the movement. The signals shown in the figure correspond to LFCs components. (B) Example of one patient’s electrodes location. (C) Example of the magnitude response of one of the learned filters showing a peak in the low-frequency domain around 1 Hz.
