Facial expression using Histogram of Oriented Gradients and Ensemble Classifier

  • Maher Kh. Hussien1
  • Fawziya Mahmood Ramo2


In this research, two methods were proposed to design a new system to recognize facial expressions. The first method relies on extracting features from the face area, and the second method relies on the process of extracting features on the parts of the face (eyes, nose, and mouth) where the histogram of oriented gradients (HOG) algorithm was used in the feature extraction process in addition to the principal component analysis algorithm to reduce feature dimensions in both methods. We have proposed a group classifier consisting of three basic classifiers: support vector machines, knn-algorithm closest to neighbors, and Naive Bayes in the classification stage. Our proposed algorithm was tested on japanese female facial expression (JAFFE) Dataset and Cohn-Kanade (CK) dataset. It was found that higher overall accuracy is achieved for F1-Score when using the second method of 93.82 % and 94.12% for CK and JAFFA, respectively.



Jun 2, 2022
How to Cite
HUSSIEN1, Maher Kh.; RAMO2, Fawziya Mahmood. Facial expression using Histogram of Oriented Gradients and Ensemble Classifier. Tikrit Journal of Pure Science, [S.l.], v. 27, n. 2, p. 52-56, june 2022. ISSN 2415-1726. Available at: <https://tjps.tu.edu.iq/index.php/j/article/view/1283>. Date accessed: 04 july 2022. doi: http://dx.doi.org/10.25130/j.v27i2.1283.