Face Identification and Verification using Hidden Markov Model with Maximum Score Approach

Objectives/Background: To analyse image using highest score estimation which will help to identify the accuracy in term of recognition rate with different likelihood. Methods/Statistical Analysis: To perform the accuracy measurement, first we initialize the HMM to find out the uniform segmentation and then extracting the feature of image by parameter initialization and subsequently train the feature image using Viterbi algorithm. Further, input image is converted into super and embedded state and transformation matrix is evaluated using maximum score estimation for each face. After that we check for maximum score HMM parameters if found ‘YES’ or go for some new feature extraction if found ‘NO’ condition. Findings: In this paper we try to find out the rate of recognition for different approaches using different class and compare it to achieve the best classification performance with twenty five facial images. Maximum score approach has been widely reported in many literature but only for the verification purpose. But, here we try to interpret the highest level of maximum score HMM by considering different class of feature of face which will also be used for identification and verification application. We also try to highlight that this algorithm brings highest process of recognition rate with massive proficiency to deal with new data and that makes it unique from other approach. Novelty/Improvement: This paper chooses discriminative feature extraction for face recognition. An experimental result concluded significant changes in term of improvement with respect to other existing conventional HMM based recognition and reflects better accuracy and efficient capability using proposed approaches. Applications: The application domain in face recognition includes surveillance, biometric, home appliances because of its richness. It is also used for access control like PC face based login, PC cameras beyond the password and physical security. Keywords: Discriminative Feature Extraction, Face Identification, Face Verification, Pattern Recognition, HMM