Repository navigation
Conversation
Record how wrist detection actually behaves, and replay synthetic traces through FallAlgorithm without changing the decision. Co-authored-by: Thomas Laure <thlaure@users.noreply.github.com>
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Résumé
Audit de la détection de chute, sans modification de l'automate. Le rapport est
docs/audit-detection-chute.md. Il sépare les faits (fichier et lignes), les interprétations et les hypothèses.Constats principaux
FallAlgorithm: accéléromètre seul à 25 Hz, quatre conditions en ET (phase basse 160 ms sous 0,35 g, impact 4,2 g, rotation du poignet 80°, immobilité continue de 2 s). Les commentaires du service etdocs/SYSTEM_OVERVIEW.mddécrivent encore un OU.CMFallDetectionManager. Les seuils du téléphone ne changent pas la décision Apple.Harness
FallReplayrejoue des CSVt_ms,ax_mps2,ay_mps2,az_mps2dans leFallAlgorithminchangé, avec le reset et le silence de 5 s du service. Métriques : sensibilité, spécificité, fausses alertes par jour. Un corpus externe se branche avec-Dfall.replay.dir=...et n'échoue pas le build.Le rapport indique de mesurer ensuite le poignet FallAllD (238 Hz, ±8 g), pas SisFall (ceinture).
Vérification
./gradlew :app:testDebugUnitTestdansapps/wear_os: vert.make quality/make testà la racine non lancés : pas de changement API, Flutter, ni de logique de détection.local.propertiesn'est pas dans le commit.