Parameters Research of Facial Emotion Detection Algorithm Based on Machine Learning

The purpose of emotional state recognition is to let computers have the ability to analyze and understand human emotions and intentions, and deeply analyze human psychological activities, so as to play a role in the fields of entertainment, education, intelligent medical treatment and so on. Differe...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:2024 Asia Pacific Conference on Innovation in Technology (APCIT) s. 1 - 6
Hlavní autori: Ting, Zhou, Yadav, Amit, Jiang, Shi Xiao, Khan, Asif
Médium: Konferenčný príspevok..
Jazyk:English
Vydavateľské údaje: IEEE 26.07.2024
Predmet:
Algorithm parameters
Emotion recognition
Face recognition
Machine learning
Machine learning algorithms
Nearest neighbor methods
Psychology
Support vector machines
Technological innovation
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:The purpose of emotional state recognition is to let computers have the ability to analyze and understand human emotions and intentions, and deeply analyze human psychological activities, so as to play a role in the fields of entertainment, education, intelligent medical treatment and so on. Different emotion recognition algorithms and different parameters of the same algorithm have different recognition effects. Among them, muscle-based feature model and 68 feature point calibration are two common face emotion recognition methods. The former uses deep learning to judge the emotional state by analyzing the movement of face muscles, while the latter calculates various feature parameters through the position relationship of 68 key points of the face, and then judges the emotional state. This paper mainly discusses the calibration method of 68 feature points. Through the use of two machine learning algorithms (KNN, SVM) and the study of different parameters in the algorithm, the influence of different parameters on the emotion recognition effect is compared and analyzed. The experiment proves that by detecting 68 key points of face, we can find the optimal parameter value in the current classification task.
DOI:10.1109/APCIT62007.2024.10673529