Higher-order Multivariable Polynomial Regression to Estimate Human Affective States.

Area of Science:

• Computational psychology
• Affective computing
• Psychophysiology

Background:

• Estimating human affective states computationally is crucial for understanding emotions.
• Existing models (linear and nonlinear) have limitations in precision and complexity.
• Physiological signals like skin conductance offer objective measures of affective states.

Purpose of the Study:

• To introduce a novel computational modeling method, higher-order multivariable polynomial regression, for estimating human affective states.
• To improve the accuracy and simplicity of affective state estimation models.
• To explore the relationship between physiological signals and affective dimensions (valence and arousal).

Main Methods:

• Utilized the International Affective Picture System to induce affective states in 30 subjects.
• Employed skin conductance as a primary physiological input variable.
• Developed and applied a higher-order multivariable polynomial regression model for prediction.

Main Results:

• Achieved high correlation coefficients for predicting affective valence (0.98) and arousal (0.96).
• Demonstrated the model's efficiency in capturing complex psychophysiological relationships.
• Provided indirect evidence for the neural origins of valence and arousal in motivational circuits.

Conclusions:

• Higher-order multivariable polynomial regression offers an accurate and simplified approach to estimating human affective states.
• The method effectively utilizes physiological signals for emotion recognition.
• This technique holds potential for advancing research in affective computing and neuroscience.