An affective sensing approach through pupil diameter processing and svm classification - biomed 2010

Area of Science:

• Human-Computer Interaction
• Affective Computing
• Biomedical Signal Processing

Background:

• Pupil diameter (PD) changes correlate with affective states.
• Pupillary light reflex (PLR) significantly impacts PD measurements.
• Accurate affective sensing requires mitigating PLR interference.

Purpose of the Study:

• To develop an affective sensing approach using PD signals.
• To minimize PLR influence on PD data using an adaptive interference canceller (AIC).
• To evaluate the effectiveness of PD-based affective sensing compared to other physiological signals.

Main Methods:

• Processing pupil diameter (PD) signals from eye gaze trackers.
• Implementing an adaptive interference canceller (AIC) with an H-infinity time-varying (HITV) algorithm to reduce PLR.
• Classifying affective states using a support vector machine (SVM) with processed PD (PMPD) features.
• Comparing PD-based classification with galvanic skin response (GSR) and blood volume pulse (BVP) signals.

Main Results:

• The AIC effectively minimized PLR-induced PD variations to an acceptable level.
• Processed MPD (PMPD) signal features enabled accurate recognition of affective states (e.g., stress).
• PD-based affective sensing showed comparable or improved efficiency when integrated with GSR and BVP data.

Conclusions:

• The proposed AIC with HITV algorithm effectively reduces PLR impact on PD signals.
• Pupil diameter processing offers a viable method for affective sensing in computer users.
• This approach enhances the accuracy and reliability of emotion detection through physiological signals.