Abstract
This work introduces a novel classifier for a P300-based speller, which, contrary to common methods, can be trained entirely unsupervisedly using an Expectation Maximization approach, eliminating the need for costly dataset collection or tedious calibration sessions. We use publicly available datasets for validation of our method and show that our unsupervised classifier performs competitively with supervised state-of-the-art spellers. Finally, we demonstrate the added value of our method in different experimental settings which reflect realistic usage situations of increasing difficulty and which would be difficult or impossible to tackle with existing supervised or adaptive methods.
Highlights
A Brain-Computer Interface (BCI) [1,2] is designed to allow direct communication between man and machine
The P300 speller has already been used by patients suffering from amytrophic lateral sclerosis [4,5], and the study performed by Vaughan et al [5] has shown that the spelling system is not limited to experiments in a laboratory but can be extended to home usage
We evaluated our method in several experimental settings, in either an offline setting in order to assess the performance w.r.t. upper bounds on the performance, or in an online setting to closely mimick realistic circumstances
Summary
A Brain-Computer Interface (BCI) [1,2] is designed to allow direct communication between man and machine. In this work we focus on the P300 speller as presented by Farwell et al [3] in 1988. This system allows people to spell words by looking at the desired character in a matrix shown on screen, enabling paralyzed patients to communicate with the outside world. Most trainable methods require data for which the ground truth is known. Recording this data is very time consuming and a lot of effort has already been put into reducing the need for labeled data. There exists no other method than the one proposed in this paper which is able to train a P300 classifier without any labeled data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
