Abstract
Childhood apraxia of speech is a genetically driven, neurodevelopmental speech sound disorder with speech deficits theorized to reflect difficulty in the spatiotemporal programming of speech movements. Therefore, this work examined how well articulatory coordination features generated from audio-estimated kinematic data distinguished speakers with childhood apraxia of speech versus non-apraxic speech sound disorder. Two correlation-based feature sets motivated by recent literature demonstrated high performance in replicated 6-fold nested cross validated studies, with no statistically significant difference between feature sets (mean AUROC = .90, σ = .04). An ablation study emphasized the importance of source-filter coordination in this population of apraxic speakers, with the source-ablated feature set performing significantly worse than the lip-ablated, the tongue-ablated, and full feature set (ΔAUROC = -.19, SE = 0.01, p < .001).
Original language | English (US) |
---|---|
Pages (from-to) | 5138-5142 |
Number of pages | 5 |
Journal | Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH |
DOIs | |
State | Published - 2024 |
Event | 25th Interspeech Conferece 2024 - Kos Island, Greece Duration: Sep 1 2024 → Sep 5 2024 |
Keywords
- articulatory coordination
- childhood apraxia of speech
- clinical speech technology
- speech inversion
ASJC Scopus subject areas
- Language and Linguistics
- Human-Computer Interaction
- Signal Processing
- Software
- Modeling and Simulation