Visual data on tongue and lips in phonemes classification

This paper describes the concept of SVM classification of the vowels and some consonants based on visual information gathered from lips shape and tongue position. Support vector describing the lips shapes was extend by the data about tongue position. To minimize the vector dimension, tongue position was classified accordingly to convention used in logopedics. The goal of the experiments is to maximize the effectiveness of classifier for single patient. The class here means possible combination of the lips shapes and tongue position for one vowel but only for selected patient. This work will try to find answer on question: Is is possible to classify phonemes using other than audio information?