Pseudo 2D Hidden Markov Model Based Face Recognition System Using Singular Values Decomposition Coefficients

A new Face Recognition (FR) system based on Singular Values Decomposition (SVD) and pseudo 2D Hidden Markov Model (P2D-HMM) is proposed in this paper. The state sequence of the pseudo 2D HMM are modeled independently which gives superior results when compared to regular 2D HMMs. As a novel point presented here, we have maintained a limited number of quantized Singular Values Decomposition (SVD) coefficients as features vectors describing blocks of face images. This makes the FR system more robust and less complex. Experiments are carried out to evaluate the proposed approach on the Olivetti Research Laboratory (ORL) face database. In order to reduce the computational complexity and the process memory consumption the images are resized to a specific dimension in the JPEG format. The system achieves a recognition rate of 99.5% which is much better than the recognition rates of the previous HMM approaches. Keywords— Face Recognition (FR), Singular Values Decomposition (SVD), pseudo 2D Hidden Markov Model (P2DHMM)