A Mixed Classifier Based on Combination of HMM and KNN

Abstract

Facial expression is an important communication method. Facial expression recognition has been studied in many application domains. In this paper, we study hidden Markov model (HMM) and K nearest neighbor (KNN) classifiers, and put forward a combined approach for facial expression recognition. The basic idea of this approach is to employ the HMM and KNN classifiers in a sequential way. First, the HMM classifier is used to calculate the probabilities of six expressions. From two most possible results of classification by HMM, the KNN classifier is used to make a final decision while the difference between the maximum probability and the second is less than the threshold obtained from HMM and training samples. The experiments show that the performance of this method exceeds that of solely HMM-based or KNN-based method.