Mispronunciation Detection in Articulation Points of Arabic Letters using Machine Learning

Mispronunciation detection in the articulation points of Arabic letters is a comparably less researched area. However, recently it is gaining popularity as the manual method of learning Al-Quran requires more time and manpower. In this technological era, most people, especially the younger generation, do not have time for the face-to-face learning method, where the listening, correction, and repetition of correct pronunciation take place in real-time. The system presented in this work has the capability to act as an alternative to support the manual learning process as well. Without denying the main role of Quran teachers, this system is not intended to replace the manual teaching method but to complement it and to ensure the art of reciting the Al-Quran is not lost and every person is able to somehow correct their tajweed. The mispronunciation detection system has been developed and tested to present the easiest way for Muslims to learn and recite Al-Quran with a better understanding of Tajweed rules. For feature extraction, RASTA PLP is used. The Hidden Markov Model (HMM) is used for training and recognition processes. Since the Arabic language is a very delicate language, mispronunciation detection is very challenging. However, promising results are achieved using the presented application. RASTA-PLP showed an 85%recognition rate. A real time application named Intelligent Quran Recitation Assistant (IQRA) showed 90% of the results, while a recorded application showed 98% of results. These results indicate that the conducted research is successfully implemented.