Objectives:
UX and emotions are increasingly popular field of study in HCI.
A new trend in this field is the use of physiological measurements to aid evaluating UX.
In this project, two studies investigate whether physiological measurements can be used to predict SAM ratings, and the nature of the relations of physiological measurements taken during system interaction and then again during a cued-recall session.
Methods:
Emotiv Epoc, Mindplace Thoughtstream, Arduino Pulse Sensor and Microsoft Kinect were used to collect EEG, EDA, HR and Facial data.
In the first paper, this data was used along with SAM ratings in order to train a SVM to predict the SAM values.
In the second paper, the data was collected for a number of groups both during system interaction, and during a recall session, with differing intermediate time delay and subjection to stimuli.
This data was then compared using Pearson product-moment correlation and ANOVA.
Results:
The results from the first paper confirmed that using physiological data to predict SAM values was significantly better than naively guessing.
Furthermore, it was confirmed that using sensor fusion can significantly increase the prediction accuracy.
The results from the second paper confirmed a significant relation between data collected during system interaction and during cued-recall for EEG and EDA.
Furthermore a significant decrease in correlation was found for EEG data, for larger intermediate time delays.
Conclusion:
We found high accuracy results in predicting the SAM ratings, which indicates further potential for computer-assisted UX evaluation.
The results from the second paper indicates that one should be wary of intermediate time delay even when using cued-recall methods.