An SSVEP-Based Brain–Computer Interface for Text Spelling With Adaptive Queries That Maximize Information Gain Rates

Abstract

This paper presents a brain-computer interface for text entry using steady-state visually evoked potentials (SSVEP). Like other SSVEP-based spellers, ours identifies the desired input character by posing questions (or queries) to users through a visual interface. Each query defines a mapping from possible characters to steady-state stimuli. The user responds by attending to one of these stimuli. Unlike other SSVEP-based spellers, ours chooses from a much larger pool of possible queries-on the order of ten thousand instead of ten. The larger query pool allows our speller to adapt more effectively to the inherent structure of what is being typed and to the input performance of the user, both of which make certain queries provide more information than others. In particular, our speller chooses queries from this pool that maximize the amount of information to be received per unit of time, a measure of mutual information that we call information gain rate. To validate our interface, we compared it with two other state-of-the-art SSVEP-based spellers, which were re-implemented to use the same input mechanism. Results showed that our interface, with the larger query pool, allowed users to spell multiple-word texts nearly twice as fast as they could with the compared spellers.

The performance of 9–11-year-old children using an SSVEP-based BCI for target selection

“…Non-invasive brain-computer interfaces (BCIs) enable users to control external devices (such as computer systems [1, 2], prosthetic arms [3, 4], and robotic vehicles [5]) using brain activity. The primary application of BCIs is to replace [6] function in those with severe motor disabilities that are the result of injury [7] or disease [8, 9].…”

“…Accuracy is the proportion of times the predicted target matches the target [17]. Latency is the mean time from target onset to classification [18, 2]. Bitrate is the number of bits per second that are transmitted, and is often preferred over other measures because it accounts for both accuracy and latency.…”

Superposition model for Steady State Visually Evoked Potentials

“…The SSVEP signals are used in Brain Machine Interfaces (BMI) in applications such as teleoperation control of an exoskeleton robot [1], wheelchair movement control [2], telepresence control of humanoid robot [3], speller system [4], among others.…”

Recursive Bayesian Coding for BCIs

“… 2 Since our paper was submitted for review, similar to our approach, another paper which describes an adaptive query strategy for a discrete memoryless channel has been published [36]. However, there is an important fundamental difference between the approaches of these two papers: unlike the above mentioned published work, our optimization searches over all possible code vectors (1) without requiring our task or brain symbols to be in any particular order.…”