Abstract— Speech recognition technology is used in learning to read letters in the Qur'an. This study aims to implement the CNN algorithm in recognizing the results of introducing the pronunciation of the hijaiyah letters. The pronunciation sound is extracted using the Mel-frequency cepstral coefficients (MFCC) model and then classified using a deep learning model with the CNN algorithm. This system was developed using the CRISP-DM model. Based on the results of testing 616 voice data of 28 hijaiyah letters, the best value was obtained for accuracy of 62.45%, precision of 75%, recall of 50% and f1-score of 58%.