Abstract

Emotions are primarily thought of as mental experiences of body states, which are mostly shown in the face with precise and specific muscle patterns. It is, perhaps, the most critical attribute of living beings, and is extremely difficult to detect and generate artificially. Its detection always remains a well-explored classical problem. Existing approaches for detecting human emotions generally demand significant infrastructural overheads. Excluding these overheads, in this paper, we propose a much simpler way of emotion detection. To do so, We have induced different states of emotion through different multimedia components, and then collected participants' keystrokes (free text) and mouse usage data through a custom-developed survey. We have used several existing classifiers (KNN, KStar, RandomCommittee and RandomForest) and a newly proposed light-weight classifier namely Bounded K-means Clustering, to analyze those usage data for different emotional states. Our analysis demonstrates that emotion can be detected from the usage data up to a certain level. Moreover, our proposed classifier enables the best detection of five emotional states namely happiness, inspiration, sympathy, disgust, and fear compared to other existing classifiers. Besides, the analysis also reveals that user identification through usage dynamics does not result in a good level of accuracy when usage gets influenced by different emotional states.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call