Phoneme recognition using neural networks

• Main Authors: Aquino, Robert Roveriskis M., Cheng, Christian Calvin T., Rule, Dionisio C.
• Format: text
• Language: English
• Published: Animo Repository 1995
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/10945
• Institution: De La Salle University

Understanding speech has always been among the few things that the computer is capable of doing. This is probably because understanding speech involves so many steps which are not clear-cut. We as persons understand speech easily, but we do not understand how we actually do it. The most fundamental aspect of recognizing speech - understanding which sounds the utterances are making - is already very difficult to simulate on a computer. Traditional approaches to this problem has always been to extract parameters which maybe useful in classifying the sounds, for example, spectral parameters, and then using a lot of statistics and algorithms to classify the different sounds. This thesis will show that there is an easier way to identify sounds into distinct phonemes. Instead of using statistics and different algorithms to classify phonemes, neural networks will be used. It will be seen that its implementation would be much simpler and results similar to, if not better, than the results obtained from using traditional methods.