Human stress classification during public speaking using physiological signals.

Area of Science:

• Physiology
• Biomedical Engineering
• Neuroscience

Background:

• Public speaking induces stress, impacting health.
• Accurate stress detection is vital for intervention.
• Multimodal physiological signals offer potential for stress assessment.

Purpose of the Study:

• To propose a multimodal classification scheme for public speaking stress.
• To evaluate the efficacy of fused physiological signals in stress detection.
• To identify significant physiological markers of stress during public speaking.

Main Methods:

• Acquired electroencephalography (EEG), galvanic skin response (GSR), and photoplethysmography (PPG) signals from 40 participants.
• Extracted frequency-domain EEG features and time-domain GSR/PPG features.
• Utilized a leave-one-out cross-validation with five classifiers, including support vector machine (SVM).

Main Results:

• Achieved a highest classification accuracy of 96.25% using an SVM with radial basis function.
• Demonstrated statistically significant differences in EEG, GSR, and PPG signals between stressed and non-stressed states.
• Confirmed the credibility of brain activity, skin conductance, and blood flow as stress indicators.

Conclusions:

• The proposed multimodal approach accurately classifies public speaking stress.
• EEG, GSR, and PPG are reliable physiological measures for detecting stress.
• This method can inform strategies to mitigate the health impacts of public speaking stress.