Image preprocessing techniques for facial expression classification
Andre Luiz da S. Pereira, Leandro A. F. Fernandes, Aura Conci · 2022
Facial expression classification is an essential aspect of human interaction, and its proper classification is considered fundamental for the future inclusion of multimedia psychology into human-computer interfaces. For this purpose, this article aims to identify a preprocessing pipeline capable of reduce the accuracy variance for facial expression classification. Thereunto, Extended Cohn-Kanade Dataset, Support Vector Machine, and Bag of Features were used in six pipelines. Compared to a pipeline with minimal preprocessing, the best results presented an accuracy variance reduction of 65.63%.