This work focused on applying the Convolutional Neural Networks (CNN) to recognize one recitation rule of the Holy Quran, the Qalqala recitation rule which is applied to letters (Ba/Dal/Jem/Qaf/Ta) of the Arabic Alphabet and it implies vibration of these letters when there is absence of vowels on them with sukun. The feature extraction technique used in the suggested system was the Mel Frequency Cepstral Coefficients (MFCC) its output was pre-processed then fed to the CNN as input to start the recognition process. Recognition process consists of two stages, the first stage was assigned with letter identification and it achieved 92% accuracy, the second stage was responsible of recognizing whether or not the identified letter is in Qalqalah status, it scored 99% for Baa, 93% for Daal, 95% for Jeem, 92% for Qaaf and 83% for Taa. The above mentioned Alphabet with sukun were used as the main dataset and they were annotated out of continuous audio signals for professional reader Ayman Swayed, each sample represent one of the Qalqalah letters with length of 300ms.
Omran, D., Kandil, A., ElBialy, A., Samy, S., & fawzy, S. (2023). CNN for speech recognition case study: Recitation Rules of the holy Quran. MSA Engineering Journal, 2(4), 1-12. doi: 10.21608/msaeng.2023.225120.1335
MLA
Dahlia Mohammad Omran; Ahmed Hisham Kandil; Ahmed ElBialy; Sherif Samy; Sahar fawzy. "CNN for speech recognition case study: Recitation Rules of the holy Quran", MSA Engineering Journal, 2, 4, 2023, 1-12. doi: 10.21608/msaeng.2023.225120.1335
HARVARD
Omran, D., Kandil, A., ElBialy, A., Samy, S., fawzy, S. (2023). 'CNN for speech recognition case study: Recitation Rules of the holy Quran', MSA Engineering Journal, 2(4), pp. 1-12. doi: 10.21608/msaeng.2023.225120.1335
VANCOUVER
Omran, D., Kandil, A., ElBialy, A., Samy, S., fawzy, S. CNN for speech recognition case study: Recitation Rules of the holy Quran. MSA Engineering Journal, 2023; 2(4): 1-12. doi: 10.21608/msaeng.2023.225120.1335
)
View on SCiNiTO