Study of VTLN method to recognize common speech disorders in speech therapy of Persian children

In this study, isolated word recognition system has been investigated. The purpose of this research is to increase the performance of children´s speech recognizer using vocal tract length normalization. The speech data that has been used to recognize phase, are correct and incorrect pronunciation of 47 words. This recognition system has been created to design a speech therapy software. Recognition of correct and wrong pronunciation and help children to improve it are the goals of this software. Four Baseline models have been Trained, one for children, one combined model (females and children) and two for Adults (by exploiting one Persian database). Children´s model was trained with data that has been collected from 25 children. 13 children´s vocabulary data were used to recognize phase. These experiments were implemented in HTK toolkit. Poor performance was improved using VTLN. Improvement of adult´s model was more than children´s model.