A Novel Facial Feature Extraction Method Based on ICM Network for Affective Recognition

Abstract

This paper presents a facial expression recognition<br><br>approach to recognize the affective states. Feature extraction is<br><br>a vital step in the recognition of facial expressions. In this work,<br><br>a novel facial feature extraction method based on Intersecting<br><br>Cortical Model (ICM) is proposed. The ICM network which is<br><br>a simplified model of Pulse-Coupled Neural Network (PCNN)<br><br>model has great potential to perform pixel grouping. In the<br><br>proposed method the normalized face image is segmented into<br><br>two regions including mouth, eyes using fuzzy c-means<br><br>clustering (FCM). Segmented face images are imported into an<br><br>ICM network with 300 iteration number and pulse image<br><br>produced by the ICM network is chosen as the face code, then<br><br>the support vector machine (SVM) is trained for discrimination<br><br>of different expressions to distinguish the different affective<br><br>states. In order to evaluate the performance of the proposed<br><br>algorithm, the face image dataset is constructed and the<br><br>proposed algorithm is used to classify seven basic expressions<br><br>including happiness, sadness, fear, anger, surprise and hate<br><br>The experimental results confirm that ICM network has great<br><br>potential for facial feature extraction and the proposed method<br><br>for human affective recognition is promising. Fast feature<br><br>extraction is the most advantage of this method which can be<br><br>useful for real world application.