Application of neural networks for the detection of pathological articulation for Serbian phonemes

Bearing in mind the existential significance of speech there is a well-developed approach to the pathology of speech that is used in speech therapy practice. Based on that, some important theoretical principles of a software package for automatic categorization and quantification of the speech production are discussed. This paper presents the results of the artificial neural networks application to the detection of articulation disorders at the level of phoneme. An overview of the results for different topologies of neural networks as well as the various input parameters are shown.