Combined support vector machines and hidden Markov models for modeling facial action temporal dynamics

摘要: The analysis of facial expression temporal dynamics is great importance for many real-world applications. Being able to automatically analyse muscle actions (Action Units, AUs) in terms recognising their neutral, onset, apex and offset phases would greatly benefit application areas as diverse medicine, gaming security. base system this paper uses Support Vector Machines (SVMs) a set simple geometrical features derived from detected tracked feature point data segment action into its phases. We propose here two methods improve on classification accuracy. first technique describes the original time-independent over period time using polynomial parametrisation. second replaces SVM with hybrid SVM/Hidden Markov Model (HMM) classifier model classifier. Our results show that both techniques contribute an improved Modeling by SVM-HMM attained statistically significant increase recall precision 4.5% 7.0%, respectively.