Entropy-Dominated Temporal Vocal Dynamics as Digital Biomarkers for Depression Detection

arXiv cs.LG / 5/1/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The study evaluates whether entropy-based temporal voice biomarkers can detect depression more effectively than standard static, pooled conversational features.
Using the DAIC-WOZ dataset with 142 labeled participants, the researchers reconstruct utterance-level acoustic trajectories and compare multiple temporal-feature approaches under leakage-aware validation.
Static pooling yields an AUC of 0.593, while modeling trajectory dynamics improves it to 0.637, and entropy biomarkers provide the best statistically significant gain with AUC 0.646 (nested AUC 0.615; permutation p = 0.017).
Entropy biomarkers outperform alternative temporal measures such as recurrence quantification, coupling biomarkers, sample entropy, and fractal-based features, with several biomarkers showing stability across validation folds.
The results suggest depression-relevant information may be captured more by the uncertainty/entropy of conversational dynamics than by average acoustic levels, enabling more temporally informed digital phenotypes for mental-health assessment.

Abstract

Automated depression detection often relies on static aggregation of conversational signals, potentially obscuring clinically meaningful behavioral dynamics. We investigated whether entropy-driven temporal biomarkers improve depression detection beyond standard pooled features using the DAIC-WOZ corpus. Using 142 labeled participants, we reconstructed utterance-level acoustic trajectories and compared pooled temporal baselines, trajectory dynamics, Shannon entropy biomarkers, recurrence quantification, sample entropy, fractal complexity, and coupling biomarkers under leakage-aware validation. Static pooling achieved an AUC of 0.593, trajectory dynamics improved performance to 0.637, and entropy biomarkers produced the strongest statistically significant improvement over pooled baselines (AUC 0.646; nested cross-validated AUC 0.615; permutation p = 0.017). Entropy biomarkers outperformed recurrence, coupling, sample entropy, and fractalbased features, with several biomarkers stable across folds. These findings suggest depression-related signal may lie less in average acoustic levels than in entropy of conversational dynamics, supporting temporally informed digital phenotypes for mental-health assessment.