Utilizing a Language Model to Improve Online Dynamic Data Collection in P300 Spellers

Abstract

P300 spellers provide a means of communication for individuals with severe physical limitations, especially those with locked-in syndrome, such as amyotrophic lateral sclerosis. However, P300 speller use is still limited by relatively low communication rates due to the multiple data measurements that are required to improve the signal-to-noise ratio of event-related potentials for increased accuracy. Therefore, the amount of data collection has competing effects on accuracy and spelling speed. Adaptively varying the amount of data collection prior to character selection has been shown to improve spelling accuracy and speed. The goal of this study was to optimize a previously developed dynamic stopping algorithm that uses a Bayesian approach to control data collection by incorporating a priori knowledge via a language model. Participants ( n = 17) completed online spelling tasks using the dynamic stopping algorithm, with and without a language model. The addition of the language model resulted in improved participant performance from a mean theoretical bit rate of 46.12 bits/min at 88.89% accuracy to 54.42 bits/min ( ) at 90.36% accuracy.

Subject-Independent P300 BCI Using Ensemble Classifier, Dynamic Stopping and Adaptive Learning

“…The confidence level checking process is implemented continually after each iteration, and is almost based on the Bayesian updating scheme [1], [10], [11], [20]- [22]. However, since those schemes do not work with our adaptive SVM, we propose a method which based on the SVM scores.…”

“…The other works which were mentioned in the literature [1], [10], [11], [21], [30] were not implemented on the Akimpech dataset, as well as not on the subject-independent basis. However, they all have the same dynamic stopping method of Bayesian approach, which was first proposed by [30].…”

“…To deal with this trade-off, various dynamic stopping (DS) methods had been proposed [1], [10], [18]- [22] to abruptly stop the flashing paradigm at any moment, when a set of user-defined conditions is satisfied. DS methods therefore can significantly reduce the spelling time while also preserving the expected accuracy that could be achieved by static stopping methods.…”

“…Mainsah [1], [10], Clements [21], Kindermans [11], [22] all used the similar DS method on different signal processing frameworks. They implement the DS by updating the posterior probabilities of each letter via a Bayesian interference scheme after each iteration, given the classifier score history of the previous iterations.…”

Using the detectability index to predict P300 speller performance

“…The BCI system has the flexibility to collect more or less data depending on the current signal SNR quality. Dynamic stopping algorithms have been shown to improve performance when compared to static stopping [12, 13, 14, 15, 16, 17, 18, 19], with positive preliminary online results in target BCI end-users [7]. For practical reasons, a data collection limit is usually imposed when using a dynamic stopping algorithm as convergence to the stopping threshold is not always guaranteed within a reasonable amount of time.…”