Abstract
A statistical approach of the automatic segmentation of the speech signal is discussed. It differs from more classical approaches of speech recognition systems which use techniques of artificial intelligence. The idea is that each stationary unit of the signal can be modeled by a statistical model (autoregressive model AR) and that a sequential detection of abrupt changes in the parameters of these models can be done with a test statistics.
The three simple on-line procedures which are proposed differ in the nature of the excitation of the model (the glottal impulsion can be taken into account or not) and in the nature of the test statistics (generalized likelihood, statistics of cumulative sum type).
Starting from these basic procedures, a final forward and backward strategy of automatic segmentation is designed and presented which gives better results.
We expect that this method is speaker independent. Furthermore, a good parametrisation of each segment is obtained as a byproduct.
Preview
Unable to display preview. Download preview PDF.
Bibliography
M. BASSEVILLE: "The two models Approach for the On line Detection of Changes in A.R. processes". In this book, chap. 6.
M. BASSEVILLE, A. BENVENISTE: "Détection séquentielle de changements brusques des caractéristiques spectrales d'un signal numérique". Rapport IRISA n o 161/INRIA n o 129, avril 1982.
M. BASSEVILLE, A. BENVENISTE: "Sequential detection of Abrupt Changes in Spectral Characteristics of Digital Signals". IEEE Trans. on Information Theory, Vol.29, n15, p. 709–723.
A. BENVENISTE: "Algorithmes simples d'estimation en treillis pour les séries longues". Outils et modèles mathématiques pour l'automatique, l'analyse des systèmes et le traitement de signal. Vol.2, C.N.R.S. Ed. 1982.
A. BRANDT: "Modellierung von Signalen mit sprunghaft veränderlichem Leistungsspektrum durch adaptive Segmentierung". Dissertation, MĂ¼nchen 1984.
J. CAELEN, G. PERENNOU: "Indices et Traits acoustiques dans un système de reconnaissance de la parole continue, quelques résultats". GALF, 9ème J.E.P., June 1978, Lannion.
J. DESHAYES, D. PICARD: "Off-time Statistical Analysis of Model Changes Using likelihood Methods or Non Parametric Techniques". in this book, chap. 5.
R.J. FONTANA, R.M. GRAY, J.C. KIEFFER: "Asymptotically Mean Stationary Channels". IEEE Trans. on Information Theory, Vol.II 27, n o 3, May 1981.
J.P. HATON, M. LAZREK: "Segmentation et Identification des Phonèmes dans un système de Reconnaissance automatique de la Parole Continue". 4ème Congnès A.F.C.E.T. Reconnaissance des Formes et Intelligence Antibicielle. January 1984. Paris.
O.H. KLATT: "Scriber and Lafs: two new approaches to speech analysis". Trends in speech necognition, Wayne A. Lea, Prentice Hall.
B. LOWERRE: "The Harpy speech understanding system". Trends in speech necognition, Wayne A. Lea, Prentice Hall.
J.O. MARKEL, A.H. GRAY: "Linear Prediction of Speech". Speech". Springer-Verlag N.Y., 1976.
M. MORF: "Recurisve least Squares Ladder Estimation algorithms". IEEE Trans. on ASSP, Vol. ASSP-29, n o 3, June 1981.
J. SEGEN: "Pattern-Directed Signal Analysis: Unsupervised Model Inference, applications to E.E.G. and Speech". Thesis, Carnegy-Mallon University. March 1980.
Editor information
Rights and permissions
Copyright information
© 1985 Springer-Verlag
About this paper
Cite this paper
Andre-Obrecht, R. (1985). On line segmentation of speech signals without prior recognition. In: Basseville, M., Benveniste, A. (eds) Detection of Abrupt Changes in Signals and Dynamical Systems. Lecture Notes in Control and Information Sciences, vol 77. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0006398
Download citation
DOI: https://doi.org/10.1007/BFb0006398
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-16043-4
Online ISBN: 978-3-540-39726-7
eBook Packages: Springer Book Archive