PLoS ONE : accelerating the publication of peer-reviewed science

| Next

View Larger Image Download original TIFF Download PowerPoint Friendly Image

Figure 1
Sessions show a standard BCI procedure: spatial filter and classifiers are learned each session anew from a calibration recording (e.g. with CSP and LDA) before they are applied during a feedback application.

Figure 1. Sessions 1 to N-1 show a standard BCI procedure: spatial filter and classifiers are learned each session anew from a calibration recording (e.g. with CSP and LDA) before they are applied during a feedback application.

The new Zero-Training method eliminates the calibration recording: spatial filters and a classifier are predetermined before session N starts. The spatial filters for session N are extracted from old spatial filters (blue), the classifier for session N is calculated from old calibration recordings (red). The feedback application of session N is preceded only by a very quick bias adaptation (yellow).