Source record

Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2609.21436.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

AESSI is an around-ear electrophysiological silent speech interface that uses masked-context representation pretraining to generalize across recording days. In tests with 24 participants, it reached 92.24% mean accuracy without test-day calibration and 98.0% in 21-day-later online tests, exceeding baselines by 50+ percentage points.

Concrete numbers, code, and a demo for a wearable speech interface that holds up across days without recalibration, relevant to hands-free input pipelines and accessibility tooling.

Cited by

No citations on record.