Attended, journal club type class
- Brain–Computer Interface Translates Signals Into Speech With 97% Accuracy | Technology Networks
- An Accurate and Rapidly Calibrating Speech Neuroprosthesis | New England Journal of Medicine
Patient:
- 45 year old man with ALS, symptoms 5 years before study
- ALS Functional Rating Scale–Revised score of 23 (range 0-48, higher better)
- modified Mini–Mental State Examination score of 27 (scores range from 0 to 27, with higher scores being consistent with better cognitive function)
- Communication:
- When speaking to expert listeners, he communicated at a mean (±SD) rate of 6.8±5.6 correct words per minute (the rate of conversational English is approximately 160 words per minute)
- Mean typing: 6.3±1.3 correct words per minute
Paper:
- Use microelectrode array in ventral premotor (6v), M1(4), Middle Precentral Gyrus (55b)
- Decode with nerual net, outputs phoneme probability, decoded by 5-gram model and then transformer.
- This is displayed on the screen and by a personalized text-to speech software
Achieves HIGH accuracy on full dictionary dataset