A Classification Method of Children withDevelopmental Dysphasia Based on Disorder Speech Analysis

摘要

This paper focuses on method developed for classifications of the speech with disorders. We are interested in children's neurological disorder called developmental dysphasia and its influence on the speech. Described classification method is based on children's speech signal analysis and allows observing the trend of the speech disorder during therapy. The classification is based on the fact that the disorder has an direct impact on speech production (i.e. movement of vocal tract). Thus, we can measure the trend of the disorders comparing patterns obtained from speech of healthy children to the patterns obtained from children with disorder. Described method is based on cluster analysis of Koho-nen Self-Organizing Maps (KSOMs) trained on speech signals. The main advantage of using artificial neural network is adaptability to specific attributes of the signal and tolerance for the noise contained in recordings.