Automatic detection for some common pronunciation mistakes applied to chosen Quran sounds

This paper describes an automatic system for the detection of some common pronunciation mistakes occurring in Quran recitation. It addresses the application of the Arabic language pronunciation rules. The system is a basic step towards a complete automatic teaching system of the Holy Quran recitation rules. The focus of this study is to detect the non proper pronunciation of a chosen set of emphatic and non-emphatic uttered letters. To achieve this goal, first we have developed several algorithms to extract the discriminating feature from specific uttered words that include the letters of interest. Then we developed a new semi automatic approach for segmentation of the considered speech to extract the target sounds. The reference in this study is the well recognized recitation rules of the Holy Quran. A database of correct readings for the chosen uttered letters was recorded from six famous referenced readers to be the control set of our system. The experiments are applied to test the pronunciation accuracy of the system users. The results showed the success of the system by 100% for the samples taken to detect the correct pronunciations and most common predictable mistakes.