|
ИСТИНА |
Войти в систему Регистрация |
ИСТИНА ЦЭМИ РАН |
||
The Advantage of the Methodology for Direct Evaluation of Facial Expressionsстатья
Дата последнего поиска статьи во внешних источниках: 20 октября 2021 г.
- Авторы: Gusev Alexey, Baev Mikhail, Kremlev Alexander
- Журнал: International Journal of Psychophysiology
- Том: 168
- Номер: Supplement
- Год издания: 2021
- Издательство: Elsevier BV
- Местоположение издательства: Netherlands
- Первая страница: 151
- Последняя страница: 151
- DOI: 10.1016/ijpsychj.2021.07.426
- Аннотация: Background: The purpose of our research is to develop an automatic facial expression analysis methodology and technology. Currently there are no reliable computer FACS (cFACS) tools for facial activity evaluation. We had developed our proprietary cFACS based F2F Emotion Studio software.Methods: The author's computer vision procedures for modeling of perceptive activity of FACS coding experts had been used. Empirical testing of the accuracy and reliability of developed software was carried out using the samples from SAMM and CASME II databases.Results: There are two main approaches for facial expressions automated analysis: selective and comprehensive. Selective analysis assumes search for matching facial expressions to samples from a set of 6-7 basic emotions. This approach main shortcomings are:1.Fast facial movements are almost undetectable.2.It is impossible to assess emotions by facial expression among 7-10% of population.3.Individual features of facial activity require manual adjustment of the emotion detector response levels.4.Racially biased.5.The action units (AUs) of the FACS system are detected by a neural network trained on a limited dataset.The comprehensive analysis is based on a detection of AUs combination over time. Its advantages are:1.The use of the FACS system allows to describe all possible facial expressions and distinguish emotions from conversational signals, mimic habits etc.2.Assessment of individual variability of facial movements, for example, facial hyperkinesis. No special system configuration is required.3.The possibility of complex analysis of the face with the identification of temporary patterns of facial movements.4.Completely nonracially biased.We realized the second approach advantages when creating the F2F Emotion Studio software. Five computer vision procedures focused on the direct analysis of the face surface deformities corresponding to individual AUs were developed. These procedures allow us to model the FACS expert perception of facial surface changes. 60 videos taken from the above-mentioned databases were analyzed. More than 90% of the AUs was found. The remaining 10%, in our opinion, were incorrectly marked by FACS coding experts, which is especially typical for SAMM database. After making adjustments to the encoding of video recordings, we achieved 100% detection of all AUs.Conclusions: The ability to precisely detect the main AUs and use them to describe real emotions and other types of facial activity allows us to advance in solving the problem of objective study of both macro- and micro-facial expressions.
- Добавил в систему: Гусев Алексей Николаевич
Прикрепленные файлы
| № | Имя | Описание | Имя файла | Размер | Добавлен |
|---|---|---|---|---|---|
| 1. | Полный текст | 1-s2.0-S0167876021006322-main.pdf | 48,7 КБ | 14 сентября 2021 [angusev] | |
| 2. | Иллюстрация | Presentation.pdf | 253,6 КБ | 14 сентября 2021 [angusev] |