Modeling Madd Reading Classification in Surah Al-Fatihah with MFCC Feature Extraction and LSTM Algorithm

Currently there is no standardization that regulates the length and shortness of madd readings. This research aims to classify madd recitation in Surah Al-Fatihah using Mel Frequency Cepstral Coefficient (MFCC) feature extraction and Long Short-Term Memory (LSTM) algorithm. This research uses the Machine Learning Life Cycle method and begins with the Data Collection process in the form of recording to capture qira'ah data, then audio conversion is carried out and the results will be cut to select the madd section. MFCC feature extraction is used on the selected audio and the results are used to train the classification model using the LSTM algorithm. After that, the model was tested using Confusion Matrix, from two types of models and three scenarios of data separation ratio tested, the best performance was achieved in model b scenario 2 using a data separation ratio of 80:20 and 4 LSTM layers with 64 units. The dataset used amounted to 100 data, with 80 data for training and 20 data for testing. The model achieved training accuracy of 96.88% and testing accuracy of 90.0%. The results of this study provide information and knowledge about madd tajweed classification to readers, that madd tajweed classification has never been done in previous studies and show that the LSTM algorithm and MFCC feature extraction are suitable for use in this classification model.