An automatic speech segmentation into syllable is an important task in a modern syllable-based speech recognition. It is generally developed using a time-domain energy-based feature and a static threshold to detect a syllable boundary. The main problem is the fixed threshold should be defined exhaustively to get a high generalized accuracy. In this paper, an optimization method is proposed to adaptively find the best threshold. It optimizes the parameters of syllable speech segmentation and exploits two post-processing methods: iterative-splitting and iterative-assimilation. The optimization is carried out using three independent genetic algorithms (GAs) for three processes: boundary detection, iterative-splitting, and iterative-assimilation. Testing to an Indonesian speech dataset of 110 utterances shows that the proposed iterative-splitting with optimum parameters reduce deletion errors more than the commonly used non-iterative-splitting. The optimized iterative-assimilation is capable of removing more insertions, without over-merging, than the common noniterative-assimilation. The sequential combination of optimized iterative-splitting and optimized iterativeassimilation gives the highest accuracy with the lowest deletion and insertion errors.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.