Investigation of HTK for Arabic Phonemes Boundary Detection
2017
In this paper we propose an automatic Arabic phonemes boundary detection system.This system is mainly used to perform an automatic speech corpus labeling.Because the manual labeling is a hard task and consumes time.We have used the HTK (Hidden Markov Tools Kit) model to solve this problem.The Hidden Markov Models implementation is used to detect phonemes boundaries with the textual information given by the transcription file.The final system improves a Correct Classification Rate of 89.5%, obtained by 5-HMM of 8 Gaussian components.