DocumentCode :
1829068
Title :
A Riemannian Stopping Criterion for Unsupervised Phonetic Segmentation
Author :
Gracia Pons, Ciro ; Anguera, Xavier ; Binefa, Xavier
Author_Institution :
Dept. of Inf. & Commun. Technol., Univ. Pompeu Fabra, Barcelona, Spain
Volume :
2
fYear :
2013
fDate :
4-7 Dec. 2013
Firstpage :
294
Lastpage :
298
Abstract :
With the availability of large and heterogeneous corpora of untranscribed speech we have recently seen regained interest for algorithms to perform automatic segmentation of such data into acoustically homogeneous or phonetic units. In this paper, we face the problem of phonetic segmentation under a hierarchical clustering (HC) framework. Concretely, we focus on the task of automatically estimating the optimum number of segments in speech data. For this purpose we present aRiemannian stopping criterion that is able to automatically stop the HC processing when it is close to the underlying number of phonetic segments while providing a lower variance(robust) estimation of the optimal number of segments. We test the proposed criterion using TIMIT data and show that it outperforms previous approaches obtaining a significantly lower over/under-segmentation variance by 46, 1% relative and average improvement of 0:14 compared to a previously proposed approach. We also show that the proposed method is robust in automatically finding the correct number of segments under data source variations.
Keywords :
acoustic signal processing; learning (artificial intelligence); pattern clustering; speech processing; HC framework; Riemannian stopping criterion; TIMIT data; acoustically homogeneous units; acoustically phonetic units; automatic data segmentation; data source variations; hierarchical clustering framework; large-heterogeneous corpora; optimal segment number; over-segmentation variance; robust variance estimation; speech data; under-segmentation variance; unsupervised phonetic segmentation; untranscribed speech; Acoustic measurements; Acoustics; Clustering algorithms; Covariance matrices; Estimation; Robustness; Speech; Riemannian estimator; cluster count estimation; hierarchical clustering; speech segmentation; unsupervised learning;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Machine Learning and Applications (ICMLA), 2013 12th International Conference on
Conference_Location :
Miami, FL
Type :
conf
DOI :
10.1109/ICMLA.2013.138
Filename :
6786123
Link To Document :
بازگشت