Unsupervised optimal phoneme segmentation: theory and experimental evaluation

	Unsupervised optimal phoneme segmentation: theory and experimental evaluation
	Yu Qiao; Dean Luo; Nobuaki Minematsu
刊名	IET SIGNAL PROCESSING
	2013
英文摘要	Automatic phoneme segmentation of a speech sequence is a basic problem in speech engineering. This study investigates unsupervised phonemesegmentation without using prior information on linguistic contents and acoustic models of an input sequence. The authors formulate the unsupervisedsegmentation as an optimal problem by means of maximum likelihood, and show that the optimal segmentation corresponds to minimising the coding length of the input sequence. Under different assumptions, five different objective functions are developed, namely log determinant, rate distortion (RD), Bayesian log determinant, Mahalanobis distance and Euclidean distance objectives. The authors prove that the optimal segmentations have the transformation-invariant properties, introduce a time-constrained agglomerative clustering algorithm to find the optimal segmentations, and propose an efficient implementation of the algorithm by using integration functions. The experiments are carried out on the TIMIT database to compare the above five objective functions. The results show that RD achieves the best performance, and the proposed method outperforms the previous unsupervised segmentation methods.
收录类别	SCI
原文出处	http://www.crossref.org/iPage?doi=10.1049%2Fiet-spr.2012.0191
语种	英语
内容类型	期刊论文
源URL	[http://ir.siat.ac.cn:8080/handle/172644/4358]
专题	深圳先进技术研究院_集成所
作者单位	IET SIGNAL PROCESSING
推荐引用方式 GB/T 7714	Yu Qiao,Dean Luo,Nobuaki Minematsu. Unsupervised optimal phoneme segmentation: theory and experimental evaluation[J]. IET SIGNAL PROCESSING,2013.
APA	Yu Qiao,Dean Luo,&Nobuaki Minematsu.(2013).Unsupervised optimal phoneme segmentation: theory and experimental evaluation.IET SIGNAL PROCESSING.
MLA	Yu Qiao,et al."Unsupervised optimal phoneme segmentation: theory and experimental evaluation".IET SIGNAL PROCESSING (2013).