Peer-reviewed research
Digital Voicing of Silent Speech
David Gaddy and Dan Klein, Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020
Read the originalRead on October 17, 2026
- What it is
- A University of California, Berkeley study that converts silently mouthed words into audible speech from facial electromyography (EMG) sensors, which record muscle impulses. It introduces training on silent signals and shares a new dataset of silent and vocalized facial EMG recordings.
- How much weight it carries
- Peer-reviewed at EMNLP 2020, which gave it a best paper award. Results come from a laboratory dataset with sensors placed on the face, and the word error rates depend on the test condition, so the best figure is not a general accuracy for everyday silent speech.
Facts taken from this source
- Training on silent facial EMG reduced the transcription word error rate from 64 percent to 4 percent in one data condition of the 2020 study.
“decreasing transcription word error rate from 64% to 4% in one data condition and 88% to 68% in another” (Abstract)