Abstract
In this paper, we present a classification method of image that learning the Active Basis model of lips by Active Basis Learning, and then uses EM algorithm to determine the probability of data point clusters to classify the images which are open mouth and close mouth. Final, using EM learning do iterates between recognition and re-learn to generate the deformable template. Active Basis model uses Gabor wavelet for slightly perturbs their locations and orientations before they are linearly combined to generate the Active Basis model of lips. EM algorithm is a non-fully supervised learning, Consists of two steps that E and M. The E-step is to identify objects in the image and imputes the unknown locations, orientations, and scales based on the currently learned deformable template. The M-step then re-learns the deformable template based on the imputed locations, orientations and scales. The experiment uses 50, 100, 150, 200 lips images to train deformable template and classification, the best recognition accuracy of up to 100%. The results show that images can be effectively doing feature analysis and classification.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call