Facial expression represents an intuitive physiological response to external stimuli, and it is a crucial clue to estimate internal state of a human. Our objective is to construct a model that is capable of effectively analyzing trends of facial expression transitions along positive-negative direction for the purpose of human monitoring by observing facial expression transition in daily life. To detect and recognize arbitrary changes along positive-negative direction is still difficult with existing techniques, e.g., changes are small, motion artifacts are larger than feature points’ displacements, changes are across neutral expression, etc. To address the challenge, we propose a two-stage framework using images without detecting the locations of feature points. Initially, it involves a binary classification of input image pairs to determine expression class. Subsequently, we assess expression transition trends via a comparison network trained independently on positive and negative transitions, which allows focused analysis of the specific dynamics involved in each type of transition. The capability of the proposed framework is validated by facial expression transitions samples systematically collected by giving affective visual stimuli that evoke positive and negative facial expressions. The results demonstrate our two-stage framework outperforms conventional frameworks on dataset consists of both positive and negative facial expression transitions.
1. Medical Diagnosis
Accurately assess patients’ mental health by leveraging changes in facial expressions.
2. Enhanced Human-Computer Interaction
Elevate the user experience of intelligent systems by precisely capturing facial expressions.
3. Innovations in Security Monitoring
Utilize facial expression detection in surveillance systems to identify abnormal behaviors.
4. Emotional Interaction in Virtual Reality
Create a more authentic virtual experience through emotion-sending technology.
| 氏名 | コース | 研究室 | 役職/学年 |
|---|---|---|---|
| 張鈞垚 | プラットフォーム学卓越大学院プログラム | 中村研 | 博士2回生 |
| 下西慶 | 知能情報学コース | 中村研 | 助教 |
| 近藤一晃 | 学術情報メディアセンター | 中村研 | 准教授 |
| 中村裕一 | 学術情報メディアセンター | 中村研 | 教授 |