Identification of Quran recitation segment from speech video recording / Liliana Nulkasim @ Mohd Kassim

Identifying Quran recitation segment from speech video recording has become one of an active research themes in speech processing and in application based on Quran education. Therefore, a more efficient method for video segment identification within long speech video recording that will consuming time is urgently needed. This project develops a system to identify Quran recitation segment from speech video recording. This project applied manual video segmentation to differentiate between Quran and speech video content. This project selected 10 segmented video for Quran recitation and speech from one long speech video recording and extract the features using Praat tool. More specifically, two feature sets which are pitch and intensity are proposed to differentiate between Quran recitation and speech segment characteristics. A random forest classifier algorithm is employed in Spyder IDE using python language as a machine learning language for predict the type of an audio. The performance of the accuracy of the system will be trained and evaluated by the extracted audio features that will be compared with the segmented video which have been segmented manually. A classification accuracy of this project were 57% for pitch and 85% for intensity with the performance of 85% and 95% match accordingly. Therefore, by the accuracy of the result given has been proved that this project able to enhance the identification segment of Quran recitation.