MAKHRAJ ‘AIN PRONUNCIATION ERROR DETECTION USING MEL FREQUENCY CEPSTRAL COEFFICIENT AND MODIFIED VGG-16

Based on research conducted by the Institute of Qur'anic Sciences (IIQ) as many 65% Muslims in Indonesia are illiterate Qur'an. In previous studies, was detection Arabic word pronunciation errors against non-natives using Mel Frequency Cepstral Coefficient (MFCC) and Support Vector Machine (SVM) methods with a test result 54.6%. Due to low accuracy results this study aims design build system that can correct makhraj letter ‘ain method used is combination MFCC Convolutional Neural Network (CNN) vgg-16 structure has been modified. The dataset 1,600 voice recordings divided into two categories incorrect four variations different vowels total data 800 records each category. This several experiments CNN kernel. training model produced best all were kernels 16, 32, 64 final rate 100% for 96% validation. fathah variation, validation 94%. variation dhommah kasrah obtained 97%. Therefore, succeeded distinguishing sound measuring ‘ain. Implementing modified produces high values speech during train process.