fix(#1049): lower wakeword detection threshold from 0.80 to 0.65#1064
Merged
Conversation
The 16-frame embedding window (1.28s) is longer than the 'Hey Jandal' wake word (~0.8s). Even with the embedding ring flush fix (PR #1063), the classifier window always contains ~6/16 post-speech silence embeddings, making confidence borderline around 0.80. Lowering the high-confidence threshold to 0.65 improves recall for short wake words while the STT verification path (lowThreshold=0.50) continues to act as a false-positive guard.
Debug APK readyCommit: Updated on each push. Removed when PR is merged or closed. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
The 16-frame embedding window (1.28s) is longer than the "Hey Jandal" wake word (~0.8s). Even after the embedding ring flush fix (PR #1063), the classifier window always contains ~6/16 post-speech silence embeddings, making confidence borderline around 0.80.
Fix
Lower
WAKE_WORD_DEFAULT_THRESHOLDfrom 0.80 → 0.65.This improves recall for short wake words. The STT verification path (
lowThreshold=0.50) continues to guard against false positives — any detection in [0.50, 0.65) runs STT to confirm actual speech.Change
core/voice/src/main/java/com/kernel/ai/core/voice/WakeWordPreferences.kt— one line.Closes #1049