Thesis (PH.D.) - Cairo University - Faculty Of Engineering - Department Of Electronics and Communications

The goal of this thesis is to develop techniques based on automatic speech recognition in order to implement a Computer - Aided Pronunciation Learning (CAPL) SystemNamely our target is to build a system to teach correct recitation of The Holy Qur'an (Tajweed) The basic idea is accept user's recitation of a given verse and then assessing the quality of their recitation to produce a feedback message to help them locate their mispronounced letters and eventually overcome themTo achieve this goal , several algorithms were developed to detect errors in user recitation of the Holy Qur'anThese algorithms cover all common types of recitation errorsRobust acoustic Hidden Markov Models (HMM) of correct recitation and common recitation errors were builtA state of the art speech recognizer was implemented to detect recitation errors affecting phonetic transcriptionA phoneme duration classification algorithm was implemented to detect recitation errors related to phoneme durationsThe decision reached by the recognizer is accompanied by a confidence score to reduce effect of misleading system feedback to unpredictable speech inputs

Issued also as CD