Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Detecting Mild Cognitive Impairment Across Languages: An Analysis of Speech Features in Chinese and English

Lee Hung Wei¹, Ya-Ning Chang¹; ¹National Cheng Kung University

Presenter: Lee Hung Wei

Speech analysis offers significant potential for the early, cross-linguistic detection of Mild Cognitive Impairment (MCI), but the crucial features for this remain unclear. Our study investigated a classification model for MCI detection in both English and Chinese, using three interpretable acoustic feature sets: time-domain (TD), eGeMAPS (EGE), and short-time Fourier transform (STFT). We found that integrating multi-domain features yielded the best performance in combined language conditions. Specifically, robust cross-linguistic acoustic markers were linked to energy variation, voicing regularity, fine-grained temporal and spectral dynamics, and amplitude envelope features, as identified by group-based SHAP analysis.

Topic Area: Language & Communication

Extended Abstract: Full Text PDF