Hierarchical Self-Supervised Representation Learning for Depression Detection from Speech
This work addresses the challenge of detecting depression from speech, which is a non-invasive alternative to clinical assessments, but it is incremental as it builds on existing self-supervised learning methods.
The paper tackles the problem of speech-based depression detection by proposing HAREN-CTC, a novel architecture that integrates multi-layer self-supervised learning features with cross-attention and CTC loss, achieving state-of-the-art macro F1-scores of 0.81 on DAIC-WOZ and 0.82 on MODMA.
Speech-based depression detection (SDD) is a promising, non-invasive alternative to traditional clinical assessments. However, it remains limited by the difficulty of extracting meaningful features and capturing sparse, heterogeneous depressive cues over time. Pretrained self-supervised learning (SSL) models such as WavLM provide rich, multi-layer speech representations, yet most existing SDD methods rely only on the final layer or search for a single best-performing one. These approaches often overfit to specific datasets and fail to leverage the full hierarchical structure needed to detect subtle and persistent depression signals. To address this challenge, we propose HAREN-CTC, a novel architecture that integrates multi-layer SSL features using cross-attention within a multitask learning framework, combined with Connectionist Temporal Classification loss to handle sparse temporal supervision. HAREN-CTC comprises two key modules: a Hierarchical Adaptive Clustering module that reorganizes SSL features into complementary embeddings, and a Cross-Modal Fusion module that models inter-layer dependencies through cross-attention. The CTC objective enables alignment-aware training, allowing the model to track irregular temporal patterns of depressive speech cues. We evaluate HAREN-CTC under both an upper-bound setting with standard data splits and a generalization setting using five-fold cross-validation. The model achieves state-of-the-art macro F1-scores of 0.81 on DAIC-WOZ and 0.82 on MODMA, outperforming prior methods across both evaluation scenarios.