Emotion Observation-System (EOS)

Advisor(s)

Dr. Firas Hassan

Confirmation

1

Document Type

Poster

Location

ONU McIntosh Center; Activities Room

Start Date

24-4-2026 10:00 AM

End Date

24-4-2026 10:50 AM

Abstract

The Emotion Observation System (EO-S) is a portable, multi-sensor device designed to provide real-time emotional state estimation to assist therapists, educators, and caregivers in supporting individuals with communication difficulties. This research presents a revised validation protocol focused on optimizing an embedded 1-D Convolutional Neural Network (CNN) for the classification of physiological signals. The EO-S hardware, powered by an ESP32-S3 microcontroller, integrates a suite of non-invasive sensors including heart rate (HR), galvanic skin response (GSR), pulse oximetry (SpO2), and skin temperature (SKT).

The study employs a structured emotion induction battery to collect high-fidelity, labeled data from a target population of 100–200 adults. Participants undergo six validated psychophysiological protocols: the Maastricht Acute Stress Test (MAST) for acute stress, Proximal Anticipation for anxiety, the Sustained Attention to Response Task (SART) for boredom, and standardized media-based inductions for amusement, relaxation, and neutral baselines. Physiological data is sampled every 0.25 seconds, with the embedded CNN utilizing sliding 3–5 second windows to generate real-time estimates across the emotional spectrum.

By utilizing domain-specific induction procedures and human-subject variability, this research aims to enhance the generalizability of the EO-S machine learning module and validate its efficacy as a therapeutic intervention tool. The results will contribute to the development of more responsive, non-invasive affective computing systems in clinical and educational settings.

This document is currently not available here.

Open Access

Available to all.

Share

COinS
 
Apr 24th, 10:00 AM Apr 24th, 10:50 AM

Emotion Observation-System (EOS)

ONU McIntosh Center; Activities Room

The Emotion Observation System (EO-S) is a portable, multi-sensor device designed to provide real-time emotional state estimation to assist therapists, educators, and caregivers in supporting individuals with communication difficulties. This research presents a revised validation protocol focused on optimizing an embedded 1-D Convolutional Neural Network (CNN) for the classification of physiological signals. The EO-S hardware, powered by an ESP32-S3 microcontroller, integrates a suite of non-invasive sensors including heart rate (HR), galvanic skin response (GSR), pulse oximetry (SpO2), and skin temperature (SKT).

The study employs a structured emotion induction battery to collect high-fidelity, labeled data from a target population of 100–200 adults. Participants undergo six validated psychophysiological protocols: the Maastricht Acute Stress Test (MAST) for acute stress, Proximal Anticipation for anxiety, the Sustained Attention to Response Task (SART) for boredom, and standardized media-based inductions for amusement, relaxation, and neutral baselines. Physiological data is sampled every 0.25 seconds, with the embedded CNN utilizing sliding 3–5 second windows to generate real-time estimates across the emotional spectrum.

By utilizing domain-specific induction procedures and human-subject variability, this research aims to enhance the generalizability of the EO-S machine learning module and validate its efficacy as a therapeutic intervention tool. The results will contribute to the development of more responsive, non-invasive affective computing systems in clinical and educational settings.