Holy Quran Recitation Recognition Using NLP

Abstract

Holy Quran Recitation Learning using Natural language Processing is a researchbased project and the main goal of the project is to help Muslims to learn the Quran more efficiently. While reciting the Holy Quran, Ahkam Al-Tajweed (Quranic Recitation Rules) which are the articulation rules of the Quran must be applied properly. Various efforts were made by previous systems which were mostly based on pronunciation rules. Little effort has been made on the advanced Tajweed rules which are related to the rhythmic recitation of the Quran such as where to “prolong” and “change” certain letters. This paper addresses the problem of identifying the correct usage of the Tajweed rule in the entire Quran. Specifically, we focus on the Iqlaab rule of Tajweed faced by novice reciters. We built an in-house dataset for our problem which particularly had all the audios of the IQLAAB rule which contained both the right and wrong pronunciation of the rule. During feature extraction, we used a well-known audio processing algorithm for extracting features which is (MFCC) Mel-frequency Cepstral Coefficient (MFCC). Then we used the 2 types of algorithms which are artificial neural networks(ANN) and Long Short-Term Memory (LSTM) for classification. Our highest accuracy is 86%. This accuracy was achieved by Long Short-Term Memory (LSTM).