Hybrid feature extraction for facial emotion recognition

Autor: Ali, Hasimah, Hariharan, M., Yaacob, Sazali, Adom, Abdul Hamid
Zdroj: International Journal of Intelligent Systems Technologies and Applications; January 2014, Vol. 13 Issue: 3 p202-221, 20p
Abstrakt: In recent advances of technology, the application of facial emotion recognition for human–computer interaction (HCI) is becoming an emerging trend. This HCI depends to a large extent on its ability to recognise the facial expression and ability to withstand various kinds of noise. However, confidence in its ability in providing adequate recognition remains challenging due to subtlety, complexity, and variability of facial expression. Therefore, this paper proposes hybrid feature extraction using 2D discrete wavelet transform (DWT), principal component analysis (PCA), and linear discriminant analysis (LDA) for seven (angry, disgust, fear, happy, neutral, sad, and surprise) different facial emotion recognition from static images. The reduced features are tested using k–nearest–neighbour classifier to recognise the facial emotion. The proposed method is evaluated based on two different databases, namely Japanese female facial expression and Cohn–Kanade database. The proposed method gives promising recognition rates of 100% for JAFFE and 97.52% for Cohn–Kanade. The effect of different wavelet families on the classification performance is also investigated.
Databáze: Supplemental Index