Recognition of Emotions from Physiological Signals

The purpose of this project was to examine whether and how emotions have influence on physiological signals and to develop a method to classify certain emotions by a set of such signals.

Abstract

The purpose of this project was to examine whether and how emotions have influence on physiological signals and to develop a method to classify certain emotions by a set of such signals.

The problem

Our mission was to choose a set of biological signals that may be influenced by emotions and find a method to determine a certain emotion by those signals.

The solution

The biological signals that we chose were Galvanic Skin Response (GSR), body temperature and heart rate.
About 30 participants were examined while displaying to them different movie clips which were suspected to evoke 5 certain emotions:

  • Sadness
  • Fear
  • Entertainment
  • Surprise
  • Anger

We trained a Support Vector Machine (SVM) rather than using statistical methods. The SVM algorithm was used in three different modes:

  • “One against all”
  • “Multi class”
  • “Decision value check”

All three modes were activated in 7 different input configurations:

  • Heart rate only
  • Temperature only
  • Resistance only
  • Heart rate & Temperature
  • Heart rate & Resistance
  • Temperature & Resistance
  • Heart rate & Temperature & Resistance

 

Flow chart

1

 

Tools
For recording the participants’ signals we used the following equipment:

2

Conclusions

The smallest general and specific classification errors were achieved with the Decision value check mode while checking the participants temperature data only.

Acknowledgment

We want to thank our supervisor, Eyal, for his support and guidance. We also want thank the lab staff, Johanan Erez and Ina Krinsky for their efforts to enable us the needed environment for the project