ECoG Speech + Avatar (Metzger et al., 2023)
Measured by Metzger, Littlejohn, Silva, Moses, Anumanchipalli & Chang et al. · Nature 620 (2023)
Inputs
The measured or assumed values behind the calculations, each with its source.
- N = 1024
- ≈1,000-word general-English vocabulary for the text track (1,024-word set; a 1,152-word set was used for a separate character-rate evaluation)
- rate = 78 word/min
- Median text-decoding rate (Abstract)
- P = 0.75
- 1 − 25% median word error rate for the text track (Abstract)
Strictest ITR
Each scoring method is an upper bound on the channel, so the headline is the strictest (smallest) one for this entry. Use the score selector on the home page to view any single method across entries.
-
Error-corrected words per minute
(1 − WER) × rate = 0.75 × 78 = 58.5 net word/min
-
Shannon per-word entropy of English
H ≈ 5.0 bits/word
Credits only the information in the English produced, independent of vocabulary size.
-
Information transfer rate
58.5 word/min × 5.0 bits/word ÷ 60 s/min = 4.88 bits/s
What counts as a bit depends on the action space. The number of distinguishable actions and how likely each one is are design choices of the task, not the sensing hardware. The same modality can present a fixed set of targets, a set pruned per step by a grammar or language model, or a continuous control space. Each of these changes how many actions are live and how the probability mass is spread, and therefore the information per selection. Read the action space below before comparing headline numbers across entries.
Action space
What the user can produce at each step, and how those options are distributed.
- Structure
- Context-dependent (the live set changes per step)
- Size
- 1024 distinguishable actions
- Prior
- Context-conditioned: likelihoods depend on prior actions
- Notes
- High-density ECoG decoded to text via a neural language model over a ~1,000-word vocabulary. The live action set and word likelihoods are reweighted each step by the model, so this is not a uniform fixed-target selection. The same system also drove speech-audio and avatar outputs.
Comparability The strictest bound here is the Shannon entropy of the output text, under one predictor held constant across the whole atlas (≈1 bit per character). That shared predictor makes it directly comparable to every other text entry (keyboards, spellers, silent speech and speech BCIs) regardless of prior or vocabulary size. For most text interfaces it comes out tighter than the raw-selection bounds, but not always. Where a small vocabulary makes Wolpaw tighter, that wins instead. Any Fitts, Wolpaw or log₂(N) figure shown below is another bound on the same channel. Switch the home-page score selector to compare one across entries.
Other score types
Bounds the atlas keeps out of the default strictest headline: as-reported figures, alternate task conditions, or raw-channel ceilings that shouldn't win the headline by default. Each still carries a score type, so the home-page selector ranks this entry on it when you choose that type. Read its derivation before comparing across entries.
-
Bits per selection (Wolpaw formula)
B = log2(N) + P*log2(P) + (1-P)*log2((1-P)/(N-1)) = log2(1024) + 0.75*log2(0.75) + 0.25*log2(0.25/1023) = 6.689 bits / selection
Term 1 is the information if every choice were correct; terms 2-3 subtract the bits lost to the error rate, assumed spread evenly over the other N-1 targets.
-
Selections per second
T = 0.76923 s/selection -> 1 / 0.76923 = 1.3 selections/s
-
Information transfer rate
ITR = B * selections/s = 6.689 * 1.3 = 8.696 bits/s
-
Achieved-bitrate credit per net-correct word
N = 1,024 → log2(N − 1) = log2(1023) = 10.0 bits per net-correct selection (field-standard achieved bitrate, e.g. Webgrid; Nuyujukian 2015, which introduced the metric, used log2(N)).
-
Net-correct word rate
net-correct = 2P − 1 = 2(0.75) − 1 = 0.50 of words. At 78 word/min (0.769 s/word) → 0.50 × 78 / 60 = 0.65 correct/s.
A word error commits the wrong word rather than timing out, so incorrect = 1 − P. Same N (1,024), text-track accuracy (75%) and rate (78 wpm) as the entry's Wolpaw calc. The 1,024-word action set is reweighted each step by a neural language model, so feeding it into log2(N − 1) is a per-word capacity view, not open-vocabulary communication; the ranked figure is the 4.88 bits/s word-entropy Shannon.
-
Achieved bitrate
10.0 bits × 0.65 correct/s = 6.50 bits/s.
Source
- Authors
- Metzger, Littlejohn, Silva, Moses, Anumanchipalli & Chang et al.
- Publication
- Nature 620, 2023