State-Specific Respiratory Signatures for Affective and Stress Recognition: Interpretable Respiratory Markers, Autocorrelation Lags, and Compact CNN Models

2026-06-25 · Source: Machine Learning · Field: Science & Research — Artificial Intelligence & Machine Learning, Engineering & Applied Sciences, Health & Medical Research · Depth: Expert, quick

Summary

A study on state-specific respiratory signatures for affective and stress recognition reframes RESP-based recognition as a joint predictive and explanatory problem. Utilizing the chest respiratory channel of the WESAD dataset, the research analyzed 60-second windows with leave-one-subject-out validation. It combined compact raw-signal one-dimensional convolutional neural networks (1D-CNNs) with physically grouped handcrafted respiratory signatures. The primary application focused on binary stress versus non-stress detection, while also analyzing baseline, stress, amusement, and meditation states in a one-vs-rest setting. The feature space included respiratory timing, breath-to-breath variability, waveform statistics, spectral/time-frequency descriptors, and autocorrelation/nonlinear predictability. The raw-signal CNN model achieved 96.72% accuracy, 95.30% macro-F1, and 90.61% MCC for stress-vs-rest performance. Conversely, compact feature models excelled for baseline (MCC 65.34%), amusement (MCC 35.69%), and meditation (MCC 88.65%), demonstrating their utility for non-stress conditions.

Key takeaway

For AI Scientists developing wearable stress and affective state recognition systems, you should consider a hybrid approach. Implement raw-signal 1D-CNNs for highly accurate stress detection, given their 96.72% accuracy. Simultaneously, integrate interpretable handcrafted respiratory signatures to gain clearer insights into non-stress conditions like baseline, amusement, or meditation, where they achieved MCCs up to 88.65%. This dual strategy optimizes both predictive performance and physiological transparency in your models.

Key insights

Respiratory activity offers distinct, interpretable signatures for stress and affective state recognition, combining CNNs and handcrafted features.

Principles

• Respiratory properties differentiate affective states.
• Jointly predict and explain physiological signals.
• Raw-signal CNNs excel for stress detection.

Method

Analyzed 60s WESAD chest respiratory data using leave-one-subject-out validation. Combined 1D-CNNs on raw signals with handcrafted respiratory signatures for binary and one-vs-rest state classification.

In practice

• Use 1D-CNNs for robust stress detection.
• Apply handcrafted features for non-stress state clarity.
• Leverage autocorrelation lags (Zpm/Zmp) as markers.

Topics

• Respiratory Signatures
• Affective State Recognition
• Stress Detection
• 1D Convolutional Networks
• WESAD Dataset
• Physiological Monitoring

Best for: AI Scientist, Research Scientist

Related on AIssential

• Stress Estimation in Elderly Oncology Patients Using Visual Wearable Representations and Multi-Instance Learning — Wearable data and MIL can estimate perceived stress in oncology patients, offering continuous cardiotoxicity surveillance.
• Hybrid deep learning model for multimodal vocal and lung signal analysis in health monitoring — A hybrid deep learning model effectively integrates vocal and lung signal analysis for accurate disease detection.
• Deep Sleep Classification via EEG Signal Criticality: A Passive BCI Approach for Sleep-Improvement Neurofeedback — High-accuracy passive BCI sensing is achievable via probabilistic decoding of EEG criticality.
• End-to-End Machine Learning for Depressive State Classification via EEG and fNIRS — Biological signal analysis via EEG and fNIRS offers an objective framework for depressive state classification.
• Personalized and Context-Aware Transformer Models for Predicting Post-Intervention Physiological Responses from Wearable Sensor Data — Transformer models can predict personalized physiological responses to stress interventions using wearable sensor data.
• RSPC: A Benchmark for Modeling Stress and Psychiatric Conditions in Digitally Mediated Relationships using Psychiatrist Annotations — Modeling mental health in NLP benefits from relational context, moving beyond individual-centric approaches.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.