FIGURE 1.

Schematic of the decoding method for predicting force information from the multi-channel intracortical data. In the first stage, the raw brain signals were spatially filtered based on the proposed weighted CAR filter. The weights of common noise were adaptively estimated employing Kalman filter. Different types of information including LFP features, MUA features and the combination of LFP and MUA features were extracted from the pre-processed data in the previous stage. Obtained features were fed into a partial least square (PLS) regression model to create a force decoding model based on the different types of input information.