Computational Analysis of Stress, Depression and Engagement in Mental Health: A Survey

TL;DR

This survey addresses the gap in unified analysis of stress, depression, and engagement by compiling a taxonomy, datasets, inputs, and state-of-the-art computational approaches. It tracks the evolution from traditional ML to deep learning and advanced multimodal fusion, highlighting how datasets and modalities—from visual and physiological signals to text and motion—inform accurate detection and monitoring. Key contributions include a synthesized view of stress, depression and engagement across unimodal and multimodal data, an outline of generic processing pipelines, and a discussion of real-world applications and ethical challenges. The work underscores the practical impact of multimodal affective computing for mental health, while calling for standardized datasets, privacy-preserving learning, and context-aware, interpretable models to enable robust deployment in clinical and educational settings.

Abstract

Analysis of stress, depression and engagement is less common and more complex than that of frequently discussed emotions such as happiness, sadness, fear and anger. The importance of these psychological states has been increasingly recognized due to their implications for mental health and well-being. Stress and depression are interrelated and together they impact engagement in daily tasks, highlighting the need to explore their interplay. This survey is the first to simultaneously explore computational methods for analyzing stress, depression and engagement. We present a taxonomy and timeline of the computational approaches used to analyze them and we discuss the most commonly used datasets and input modalities, along with the categories and generic pipeline of these approaches. Subsequently, we describe state-of-the-art computational approaches, including a performance summary on the most commonly used datasets. Following this, we explore the applications of stress, depression and engagement analysis, along with the associated challenges, limitations and future research directions.

Figures (6)

• Figure 1: Illustration showing the interaction between basic emotions and psychological states: the left image shows a boy happy and engaged with his reading, while the right image shows him sad but still engaged, demonstrating that psychological states can coexist with different emotions. This image was created with https://openai.com/dall-e-2.
• Figure 2: Publication trends in stress, depression and engagement from 1996 to 2024 indicate a growing interest in these areas. This underscores the increasing importance of computational methods in addressing mental health challenges and enhancing well-being.
• Figure 3: Sample data inputs and modalities for most commonly used datasets mentioned in Section \ref{['sec:sota_results']}. Along with audio-visual modalities, they use physiological signals such as ECG, EDA and EMG and have labels for valence, arousal, dominance, trustworthiness, depression and engagement categories.
• Figure 4: Chronological emergence of computational approaches from traditional ML techniques (e.g., Naive Bayes, Logistic Regression) to advanced DL methods (e.g., GNNs, Transformers) and emerging paradigms (e.g., federated learning, cloud-edge computing), illustrating the evolving complexity and sophistication of stress, depression and engagement analysis. Here, L' denotes 'Learning'.
• Figure 5: A depiction of the generic phases used in various computational approaches for stress, depression and engagement analysis.