Continuous emotional-state inference in wearables relies on sensing the autonomic and motor correlates of affect rather than subjective labels. Leading low-power modalities are photoplethysmography (PPG) for pulse and heart rate variability, electrodermal activity (EDA/GSR) for sympathetic arousal, wrist accelerometers for context and motion artefact removal, and skin temperature for peripheral vasomotor changes. Devices such as the Empatica E4 demonstrate a practical combination of these sensors in a wrist form factor, integrating PPG, EDA, accelerometry, and temperature to support continuous monitoring. Rosalind Picard at the MIT Media Lab pioneered many of the links between these signals and affective states and frames how multimodal physiological fusion can improve detection accuracy.
Key low-power sensors
PPG is widely used because reflective photodiodes and LEDs consume little energy and provide pulse-to-pulse timing for heart rate and derived heart rate variability (HRV) metrics, which the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology identified as robust autonomic indicators. EDA sensors measure skin conductance linked to sweat-gland activity under sympathetic control; they are inherently low-power and particularly sensitive to arousal. MEMS accelerometers and gyroscopes supply low-energy context about movement and posture, enabling removal of motion artefacts and distinguishing emotion-driven physiological changes from exercise. Skin temperature sensors add information about peripheral blood flow and thermoregulatory responses, further disambiguating affective states. More detailed biosignals such as single-lead ECG, dry-electrode EMG, and wearable EEG exist but typically impose higher power and form-factor tradeoffs and are therefore less common for true continuous, day-long use.
Causes, relevance and consequences
These sensors infer emotional state indirectly by capturing autonomic nervous system outputs: sympathetic activation elevates skin conductance and reduces HRV, while parasympathetic dominance raises HRV. The relevance spans mental health monitoring, stress management, and human-computer interaction, but consequences include privacy risks and potential misinterpretation across cultures and environments. John Cacioppo at the University of Chicago emphasized that physiological responses vary by individual and social context, making baseline personalization essential. Environmental factors such as ambient temperature, humidity, and terrain affect readings, and cultural norms shape emotional expression and physiological baselines, which can bias algorithms if training data are not diverse. Responsible deployment requires transparent reporting of sensor limits, user consent, and validation against established physiological research to maintain trust and clinical relevance.