A Neural Network Architecture for Italian Vowel Recognition

A speaker independent vowel classification system will be described.Experiments are performed on the complete Italian vowel set showing a very high recognition rate on new speakers. The architecture of the vowel classifier is based on an ear model (EM) followed by a set of Multi-Layered Neural Networks (MLNNs). MLNMs are trained to learn how to recognise articulatory features like the horizontal and the vertical place of articulation related to tongue position.