A Point-Process Approach for Tracking Valence using a Respiration Belt.

Area of Science:

• Psychophysiology
• Computational Neuroscience
• Affective Computing

Background:

• Emotional valence inference is challenging due to psychological factors and subject variability.
• Emotional valence changes correlate with physiological responses, particularly in respiration signals.
• Respiration patterns offer a non-invasive window into emotional states.

Purpose of the Study:

• To develop and validate a state-space model for decoding emotional valence from respiration.
• To identify key respiratory features indicative of emotional valence shifts.
• To establish a computational framework for real-time emotional state tracking.

Main Methods:

• A binary point process model was generated using respiration features (breath depth, rate, cycle time).
• An expectation-maximization (EM) framework was employed to decode hidden valence states.
• Model parameters were optimized and validated against self-reported valence ratings.

Main Results:

• The model achieved 77% accuracy in predicting high valence events.
• The model achieved 73% accuracy in predicting low valence events.
• Key respiratory features effectively indicated underlying neural stimuli related to valence.

Conclusions:

• Respiration analysis provides a viable method for inferring emotional valence.
• The developed model demonstrates potential for long-term emotional monitoring.
• Applications include closed-loop systems and wearable devices for emotional regulation.