Video-Based Face Authentication Using Appearance Models and HMMs

Abstract-In this paper, we propose a novel face authentication been receiving broad attentions in biometrics-based identification scheme using the Active Appearance Model (AAM) and the and authentication research fields [1][2]. Hidden Markov Model (HMM). The proposed face The face recognition methods can be roughly divided into three authentication system can be divided into two parts. First, the categories: the feature-based methods [1], the template-based AAM is used to extract the low-dimensional feature vectors methods [2], and the model-based methods [3]. The feature-based including combined texture and shape information of methods are the earliest and intuitive methods. In the training phase, individual face images. The extracted feature vectors are several salient facial features are detected by using a face detector, further classified into several clusters using vector and the relation of distances between the feature points is taken as quantization. The clustered feature vectors are then the feature vector and stored in the database. In the verification characterized using HMMs to make full use of the temporal phase, the minimum distance approach is utilized to measure the information across the face images. After all parameters in the similarity between the feature vectors of the test image and the HMMs are calculated, we can dynamically determine the trained image. The template-based methods use the entire face thresholds for face authentication. An iterative algorithm is template to recognize faces and usually outperform the feature-also proposed to automatically determine a suitable number of based approaches. The model-based methods such as Active Shape HMM states and a suitable number of observation classes to Model (ASM) based and AAM-based are also popular in recen