Continuous Speech Phoneme Recognition Using Dynamic Artificial Neural Networks

Phoneme classification and recognition is the first step to large vocabulary continuous speech recognition. This step represents the acoustic modeling part of such a system. In hybrid speech recognition systems phoneme recognition is made by artificial neural networks (ANN’s). The main objective of this paper is the investigation of dynamic ANN’s, namely the Time-Delay Neural Networks (TDNN) and Recurrent Neural Networks (RNN) that are the most suitable for recognition of time sequences. There are presented two types of TDDN’s: Focused Time-Delay Neural Networks (FTDNN) and Distributed Time-Delay Neural Networks (DTDNN) respectively and a Layer Recurrent Neural Network (LRNN). The development of a phoneme recognizer application using dynamic ANN’s for OASIS Numbers databases is also described. There are also presented the phoneme classification experiments and the results for the ANN’s. Finally some conclusions are drawn based on the experimental results.