Tecnura (Jan 2024)

Emotion Recognition in the Eye Region Using Textural Features, IBP and HOG

  • Laura Jalili,
  • Josue Espejel,
  • Jair Cervantes,
  • Farid Lamont

DOI
https://doi.org/10.14483/22487638.22100
Journal volume & issue
Vol. 28, no. 79
pp. 12 – 33

Abstract

Read online

Objective: Our objective is to develop a robust emotion recognition system based on facial expressions, with a particular emphasis on two key regions: the eyes and the mouth. This paper presents a comprehensive analysis of emotion recognition achieved through the examination of various facial regions. Facial expressions serve as invaluable indicators of human emotions, with the eyes and mouth being particularly expressive areas. By focusing on these regions, we aim to accurately capture the nuances of emotional states. Methodology: The algorithm we devised not only detects facial features but also autonomously isolates the eyes and mouth regions. To enhance classification accuracy, we utilized various feature extraction and selection techniques. Subsequently, we assessed the performance of multiple classifiers, including Support Vector Machine (SVM), Logistic Regression, Bayesian Regression, and Decision Trees, to identify the most effective approach. Results: Our experimental methodology involved employing various classification techniques toassess performance across different models. Among these, SVM exhibited exceptional performance, boasting an impressive accuracy rate of 99.2 %. This outstanding result surpassed the performance of all other methods examined in our study. Through meticulous examination and experimentation, we explore the effectiveness of different facial regions in conveying emotions. Our analysis encompasses two datasets and evaluation methodologies to ensure a comprehensive understanding of emotion recognition capabilities. Conclusions: Our investigation presents compelling evidence that analyzing the eye region using a Support Vector Machine (SVM) along with textural, HoG, and LBP features achieves an outstanding accuracy rate of 99.2 %. This remarkable finding underscores the significant potential of prioritizing the eyes alone for precise emotion recognition. In doing so, it challenges the conventional approach of including the entire facial area for analysis.

Keywords