AUDIO SEGMENTATION USING SHORT-TIME ENERGY (STE) FOR MAKHRAJ RECOGNITION IN QUR' AN RECITATION WITH XGBOOST AND CONVOLUTIONAL NEURAL NETWORK (CNNLD) APPROACHES

• Main Author: Pangestu Rachman, Lingga
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/85542
• Institution: Institut Teknologi Bandung

This study proposes and evaluates an enhanced makhraj recognition system for Quranic recitation by incorporating an audio segmentation stage using the ShortTime Energy (STE) method with XGBoost and Convolutional Neural Network 1 Dimensional (CNNlD) approaches. The focus of this research is the recognition of the makhraj for the letter Lam. Makhraj refers to the articulation points in the pronunciation of Arabic letters, specifically in the context of Quranic recitation. The makhraj of the letter Lam has two classes: lam tarqiq (soft pronunciation) and lam tafkhim (heavy pronunciation), depending on the energy expended. Audio segmentation with STE has proven effective in improving the accuracy of makhraj recognition, as STE successfully identifies audio segments with the highest accumulated energy values, facilitating the voice recognition process. A onedimensional Convolutional Neural Network (CNNlD) model utilizing 39 Mel Frequency Cepstral Coefficients (MFCC) features demonstrated the most optimal results. This model indicates that the proposed approach can achieve high performance in recognizing the makhraj of the letter Lam, with an average F 1 Score of 97 .95%. The implementation of this system is expected to significantly contribute to the teaching of Quranic recitation, particularly by assisting in the accurate and efficient learning of makhraj.