Prenatal anxiety recognition model integrating multimodal physiological signal

Area of Science:

• Human-Computer Interaction (HCI)
• Maternal Health
• Affective Computing

Background:

• Anxiety in pregnant women negatively impacts well-being.
• Data scarcity and collection challenges hinder HCI interventions for pregnant individuals.
• Physiological data offers objective insights into emotional states.

Purpose of the Study:

• To develop a data-driven HCI intervention for assessing anxiety in pregnant women.
• To explore the use of wearable sensors for collecting physiological data.
• To build a machine learning model for evaluating maternal anxiety levels.

Main Methods:

• Collected physiological data (Blood Volume Pulse, Skin Temperature, Inter-Beat Interval) using Empatica E4 wristbands.
• Gathered subjective anxiety level reports from pregnant women.
• Utilized a Support Vector Machine (SVM) algorithm to model anxiety detection.
• Integrated features from multiple physiological signals.

Main Results:

• Achieved 69.3% accuracy in emotion recognition for anxiety detection.
• Demonstrated the feasibility of using wearable sensors for physiological data collection in pregnant women.
• Successfully constructed an SVM model for evaluating anxiety levels.

Conclusions:

• The developed model shows promise for HCI interventions in maternal mental health.
• Introduced conceptual biofeedback mechanisms for maternal emotion monitoring.
• Highlights potential for improved monitoring and timely interventions to support pregnant women's emotional health.