Abstract:
Voice recognition is considered as one of the most important aspects of machine learning domain. But it still has limited and modest applications in Arabic language. The Holy Quran is the largest container of Arabic language grammar in terms of speaking and utterance as it is considered as a message for all humanity. However, we present within this study a classification model for four different altajweed rules like the Allah name(mofakham, morakaq) and moon and sun L(لام),as we depended on two different kinds of voice features LPC(liner predictive coding),MFCC(Mel-frequency cepstrum), where these two types of features are the most used within the domain of processing voice signaldomain.as we depended on two classifying mechanisms (neural networks and hidden Markov model(HMM)) in order to study all possible cases of those studied rules, then we extracted those features of three different readers(males) and two different readers (female).each of Markov hidden model and neural networks have been trained by using two different types of extracted features and then we tested those trained models in order to obtain final results as to evaluate them. And resulted results in the training of Hidden Markov Models accuracy amount 90% with Allah (moufakhum), 92% with Allah (mourqeq), 83.3% with sunny لام and 80% with moony لام .
The study resulted that neural network able to distinguish between four rules of Tajweed and training samples processing with high accuracy reached 95% with Allah (moufakhum) , 94% with Allah (mourqeq) , 93% with moony لام and 92.3% with sunny
