Don't put words in my mouth: speech perception can falsely activate a brain-computer interface

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Brain-computer interfaces (BCIs) show promise for communication by decoding speech from brain activity.
• The sensorimotor cortex (SMC) is key for speech decoding BCIs but is also active during speech perception, potentially causing BCI errors.
• Understanding speech perception's interference is vital for reliable BCI control.

Purpose of the Study:

• To investigate how speech perception interferes with reliable speech BCI control.
• To determine if speech perception leads to false positive activations in speech decoding BCIs.
• To assess the distinguishability of perceived versus produced speech signals.

Main Methods:

• Recorded high-density electrocorticography (HD-ECoG) from five subjects during speech perception and production tasks.
• Trained a support-vector machine (SVM) decoder on speech production data.
• Tested the decoder on speech perception data to identify false positives and assess signal distinguishability.

Main Results:

• Both speech perception and production activate the sensorimotor cortex (SMC).
• Speech decoding BCIs trained on production data can produce false positives during speech perception.
• Perceived speech can be accurately distinguished from produced speech and rest states.

Conclusions:

• Speech perception poses a significant interference risk to reliable BCI control.
• BCI designs must incorporate strategies to mitigate false positives from speech perception for successful user adoption.
• Distinguishing perceived from produced speech is feasible, offering a pathway to improved BCI accuracy.