Hidden Markov models (HMMs) [6], which have been used successfully in speech recognition for a number of decades, are now being applied to face recognition. Samaria andYoung used image pixel values to build a top-down model of a face using HMMs. Nefian and Hayes [7] modified the approach by using discrete cosine transform (DCT) coefficients to form observation vectors. Bai and Shen [8] used discrete wavelet transform(DWT)[9]coefficientstakenfromoverlappingimage sub-windows taken from the entire face image, whereas Bicego and Murino [10] used DWT coefficients of sub-windows generated by a raster-scan of the image.