出版社:University of Malaya * Faculty of Computer Science and Information Technology
摘要:In this paper, we investigate the speech recognition system for Tajweed Rule Checking Tool. We propose a novel MelFrequency Cepstral Coefficient andVector Quantization (MFCCVQ) hybridalgorithm to help students to learn and revise proper AlQuran recitation by themselves. We describe a hybridMFCCVQ architecture toautomatically point out the mismatch between the students’recitationsandthecorrect recitationverified by the expert. The vectoralgorithm is chosen due to its data reduction capabilities and computationally efficient characteristics.We illustrate our component model and describe the MFCCVQ proceduretodevelop theTajweed Rule CheckingTool.Two features, i.e., a hybrid algorithm and solely Mel Frequency Cepstral Coefficientare compared to investigate their effect on the Tajweed Rule CheckingToolperformance. Experiments arecarried out on a dataset to demonstrate that the speed performance of a hybrid MFCCVQis86.928%, 94.495% and 64.683% faster than theMelFrequencyCepstral Coefficient for male, female and children respectively.