Automatic recognition of kinikilig laughter through body movement
Laughter is one of the most common social signals in human social interactions and is versatile enough that it could evoke a varied and complex amount of emotions. Prior literature under the field of Social Signal Processing suggests that machines could also classify laughter through various means....
Saved in:
| Main Author: | |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Animo Repository
2019
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/etd_masteral/6975 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Language: | English |
| Summary: | Laughter is one of the most common social signals in human social interactions and is versatile enough that it could evoke a varied and complex amount of emotions. Prior literature under the field of Social Signal Processing suggests that machines could also classify laughter through various means. In this research, the means of body movement was used in distinguishing the Kilig type of laughter from other types of laughter. A Kinect for Windows V2 Sensor in conjunction with the EyesWeb XMI program was used to collect the body point data and video data, MATLAB was used to extract the high-level features from the body points, and WEKA was used to run and test the classifiers as well as execute feature selection. There was one coder for this study who also annotated the videos with audio. The modelling was done using an imbalanced dataset and a balanced dataset in order to see if there was any significant change in using a balanced or imbalanced dataset. Based on the results from the modelling, the Logistic Regression algorithm had the best performance, generally. After studying the data through the numbers and visual inspection, the most significant features were found to be Head Leaning, Neck Bending, Arm Curl, Distance of Hands from Head, and the Body Lean Angle. The conclusion reached was that while body movement may be useful in classifying laughter, it would be best to include more modalities in order to add context to the body movements. |
|---|