Reckoning respiratory signals to affectively decipher mental state.

Area of Science:

• Physiological signal processing
• Cognitive science
• Brain-computer interfaces

Background:

• Recognizing mental states from physiological signals is crucial for medical diagnostics, cognitive science, behavioral studies, and brain-machine interfaces.
• Existing methods often rely on multiple physiological signals.

Purpose of the Study:

• To investigate the efficacy of using solely respiration signals for mental state recognition.
• To develop an efficient method for extracting respiratory features and addressing class imbalance.

Main Methods:

• Utilized the Affective Pacman public dataset containing physiological signals during normal and frustrated mental states.
• Implemented a novel method for removing non-linear baseline drifts to extract respiratory features.
• Applied the Synthetic Minority Oversampling TEchnique (SMOTE) to rectify class imbalance.

Main Results:

• The SMOTE algorithm significantly improved classification accuracy, reduced classifier bias, and enhanced classifier sensitivity.
• A multilayer perceptron classifier achieved 97.9% classification accuracy, 92.6% true positive rate, and 99.3% true negative rate.
• The proposed approach demonstrated superior performance compared to relevant literature.

Conclusions:

• Respiration signals alone are sufficient for accurate mental state recognition.
• The combination of advanced signal processing and SMOTE offers a robust solution for mental state classification, even with imbalanced datasets.