Friday, February 02, 2018
NoiseTag BCI - High Accuracy without Trainning
The Donders Research Group from Radboud University created a BCI/EEG platform which aims high accuracy, speed reaction and user-friendly. The main features improved are CCA-based Reconvolution, Dynamic stopping, Zero-training, Asynchronous, Headsets, Adaptive and Applications.
We found a method that turns BCI into plug and play. The first button you look at will take the system a bit longer to figure out, taking about 30 seconds. Then the second button goes down to 10 seconds. Then the 3rd-4th is down to 1-2 seconds. A person can get up to 1 button per second.
The headband uses dry electrodes, so we do not have to use water.
In the Noise-Tagging project we utilize pseudo-random noise-codes as stimulation sequences (i.e., stimuli are tagged with noise) for evoked Brain BCI. These so-called noise-tags exhibit a spread-spectrum signal and when applied as stimuli, these evoke Broad-Band Evoked Potentials (BBEP) visible in the EEG. We have designed a generative method – Reconvolution – which combines both deconvolution and convolution to learn and predict responses to these noise-tags. Specifically, adhering to the superposition hypothesis, the complex BBEP can be decomposed into a summation of time-shifted versions of a/some transient response(s).
For more information about BCI/EEG press here.