الفهرس 
Speech modules for enhancement of computer aided pronunciation learning system
SummaryOnly 14 pages are availabe for public view
 |  | from | 90 |  |  |
| | | from | 90 | | |
Abstract
Computer aided pronunciation learning (CAPL) has recently been considerable in the research community. One of the most challenging applications of a (CAPL) system is Holy Quran training for correct recitation. Gaussian mixture models (GMMS) have been the most common used models in pronunciation verification systems. The recently introduced deep neural networks (DNN) has proved to provide significantly better discriminating models of the acoustic space. In this thesis, we carried out a large number of experiments to achieve a significant improvement in the accuracy of Speech Verification system. A hybrid deep neural network-hidden markov models (DNN-HMM) approach is used for that purpose. Also, an automatic manner for selecting the training data and transcribing it was developed. As a result, we can select a sufficient part of that data with high confidence and building a very strong computer Aiding system for Holy Quran. In this work, we implemented a confidence-based scheme for automatic selection of data. This scheme is basically based on the forward backward algorithm that utilizes log-posterior probabilities. This scheme can select up to around 30% of data with accuracy reaches to 95.5%