Based Persian Viseme Clustering

Viseme (Visual Phoneme) clusterin every language is among the most important conducting various multimedia researches as reading, lip synchronization and com pronunciation training applications. With re that clustering and analyzing visemes are lan processes, we concentrated our research on P which indeed has suffered from lack of su paper, we used a hierarchical approach for c in Persian language based on principal compo polynomial kernel matrix considering coar Having obtained feature vector of each phon unweighted pair group method with arithme projected viseme on constructed manifold neighbor of the weight value as a result of rec as the criterion for comparing viseme dissimi indicate the robustness of the proposed alg experiments was conducted on Persian datab syllables were examined. Comparing the clustering algorithm with that of the percept an expert proves a reasonable evaluation algorithm. KeywordsAudio/Visual processing, C pronunciation training, Persian Viseme clu manifold.