利用中文語音聲學特徵將憂鬱症建模為漸變性狀態

// Modeling Depression as a Gradient Condition Using Acoustic Features of Mandarin Speech

本研究探討中文語音中的聲學特徵能否區分抑鬱為獨立類別或連續程度｜隨機森林分類器對話語進行判別，準確率達72.3%｜群聚分析顯示症狀呈連續分布，支持抑鬱症的維度觀點｜結果顯示語音特徵更適合反映抑鬱嚴重度，具持續監測潛力。

// This study investigated whether acoustic features of spontaneous Mandarin speech represent depression as a distinct category or a continuous severity spectrum. Using a validated speech corpus and depression scores, a random forest classifier distinguished depressed from nondepressed speech with 72.3% accuracy, while certain acoustic features correlated with depression severity. Clustering revealed overlapping groups and a continuous distribution of severity, supporting a dimensional view of depression. These results suggest speech features may better reflect symptom severity than categorical diagnosis, highlighting potential for continuous depression monitoring.

Published 2026年8月6日