Original title: A Common Measure of Communication for Speech Brain-Computer Interfaces
Dulhan Jayalath, Benjamin Ballyk, Oiwi Parker Jones · September 2026
Speech brain-computer interfacesSystems that translate brain activity into spoken language. turn brain signals into words to help people with paralysis talk. However, scientists struggle to compare these devices. Different labs test them with different brain recording tools, datasets, and word lists. A high score on a small list does not mean the device works well in daily life. To solve this, researchers created a new score called open-vocabularyReferring to systems that can handle a wide range of words, not limited to a predefined set. mutual informationA concept from information theory that measures how much knowing one variable tells you about another., or OVMI. It measures how much information a decoderThe algorithm or model that converts neural signals into text or speech. truly shares, compared to all the words a person might actually want to say. This puts different systems on one shared scale. The team discovered that standard scores like accuracy and word error rate make devices look better than they really are. Using OVMI, they revealed a trade-off: supporting more words often lowers decoding accuracy. Finally, picking words to maximize OVMI boosted decoding accuracy by up to 16.3% across three speech domains.
You're at a restaurant where the menu is completely different each time you visit. You have no way to judge if the chef can prepare the meal you're craving, because the available dishes change so often. In this paper, speech brain-computer interfaces operate with different vocabularies, like restaurants with different menus. OVMI measures how well the system communicates your intended words, regardless of which words it supports.
For people with paralysis, speech brain-computer interfaces are their voice. Without a fair way to compare systems, progress stalls, directly affecting quality of life.
The abstract doesn't specify the limitations of OVMI, but it demonstrates its application in comparing systems and improving vocabulary design.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
A Common Measure of Communication for Speech Brain-Computer Interfaces — Dulhan Jayalath, Benjamin Ballyk, Oiwi Parker Jones
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.