Speaker-dependent live quranic verses recitation recognition system using Sphinx-4 framework

This paper describes a speaker-dependent speech recognition system that recognizes and evaluates the accuracy of the recitation of the selected Quranic verses. It aims to first train the system according to the recitation of the qurra (qualified reciters) and the users themselves. The system uses an open source Carnegie Mellon University (CMU) Sphinx-4 framework based on the Hidden Markov Models (HMMs), which is one of the well-known speech recognizers used in English language. For this purpose, we have used transliteration mechanism for the Arabic language in order to generate the acoustic model. This system was tested after being trained with four different ways; Arabic word with Arabic alphabets, transliteration word with syllable, transliteration compound word with syllable, transliteration syllable with syllable. The results were promising.