Efficient Long Speech Sequence Modelling for Time-Domain Depression Level Estimation

2025-01-05

AI Summary

To address distortions in time-frequency transformations, semantic fragmentation from short-segment splitting, and insufficient modeling of depression-specific vocal dynamics in long-duration speech-based depression assessment, this paper proposes the first end-to-end time-domain modeling framework. It integrates state-space models (SSMs) with a dual-path long-sequence architecture and introduces temporal-external attention to process raw waveforms directly—fully preserving depression-relevant acoustic dynamics (e.g., prolonged pauses, slowed speech rate). By eliminating hand-crafted feature engineering and manual segmentation, the framework achieves superior performance on AVEC2013/2014, significantly outperforming state-of-the-art methods in both accuracy and robustness for long-sequence depression cue identification. It establishes a novel, interpretable, and high-fidelity time-domain paradigm for objective depression assessment in realistic conversational settings.

Technology Category

• Natural Language Processing: Speech
• Cognitive Modeling & Cognitive Systems: Affective Computing
• Machine Learning: Time-Series/Data Streams

Application Category

• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Depression significantly affects emotions, thoughts, and daily activities. Recent research indicates that speech signals contain vital cues about depression, sparking interest in audio-based deep-learning methods for estimating its severity. However, most methods rely on time-frequency representations of speech which have recently been criticized for their limitations due to the loss of information when performing time-frequency projections, e.g. Fourier transform, and Mel-scale transformation. Furthermore, segmenting real-world speech into brief intervals risks losing critical interconnections between recordings. Additionally, such an approach may not adequately reflect real-world scenarios, as individuals with depression often pause and slow down in their conversations and interactions. Building on these observations, we present an efficient method for depression level estimation using long speech signals in the time domain. The proposed method leverages a state space model coupled with the dual-path structure-based long sequence modelling module and temporal external attention module to reconstruct and enhance the detection of depression-related cues hidden in the raw audio waveforms. Experimental results on the AVEC2013 and AVEC2014 datasets show promising results in capturing consequential long-sequence depression cues and demonstrate outstanding performance over the state-of-the-art.

Problem

• Depression Detection
• Speech Analysis
• Long-duration Audio Processing

Innovation

• Depression Detection
• Long Audio Processing
• Speech Feature Extraction