Figure 3.

Open- and closed-loop user-specific decoders identification and subject training. (A) Decoder identification is performed offline by analyzing a dataset of simultaneously acquired neuronal signals and intended movements (open-loop data acquisition session). The resulting decoder is applied online on the user's neural signals so that he or she can train; that is, adapt his or her neural patterns to the imperfect decoder. Because the user progressively modifies his or her brain patterns, one or several blocks of decoder re-identification can be completed. (B) Simultaneous decoder and user adaptation using adaptive/incremental learning algorithms permits to directly identify a decoder associated with closed-loop neural patterns.