Sequence recognition with radial basis function networks: Experiments with spoken digits

Michele Ceccarelli, Joël T. Hounsou

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

In this paper we consider several learning procedures for Radial Basis Function (RBF) Networks applied to a problem of speech recognition, namely isolated word recognition. The dynamic nature of speech is considered by adding delayed connection and integration units to the network. We refer to a specific model where the layers are organised in a hirerchical manner: a first RBF hidden layer, a second sigmoidal layer and a classification layer which integrates over time the partial classifications performed by the sigmoidal layer. The training procedures for RBF networks are compared on both generalisation ability and computational costs. Our study shows that supervised learning of the centroids of the basis functions gives appreciable results at a significantly lower cost.

Original languageEnglish
Pages (from-to)75-88
Number of pages14
JournalNeurocomputing
Volume11
Issue number1
DOIs
Publication statusPublished - 1 May 1996
Externally publishedYes

Fingerprint

Radial basis function networks
Learning
Costs and Cost Analysis
Aptitude
Supervised learning
Speech recognition
Costs
Experiments
Recognition (Psychology)
Generalization (Psychology)

Keywords

  • Backpropagation
  • Isolated word recognition
  • Multilayer perceptrons
  • Neural networks
  • Radial basis functions
  • Speech processing

ASJC Scopus subject areas

  • Artificial Intelligence
  • Cellular and Molecular Neuroscience

Cite this

Sequence recognition with radial basis function networks : Experiments with spoken digits. / Ceccarelli, Michele; Hounsou, Joël T.

In: Neurocomputing, Vol. 11, No. 1, 01.05.1996, p. 75-88.

Research output: Contribution to journalArticle

@article{ee18c19696194eb99b7ec0d3a3266e7b,
title = "Sequence recognition with radial basis function networks: Experiments with spoken digits",
abstract = "In this paper we consider several learning procedures for Radial Basis Function (RBF) Networks applied to a problem of speech recognition, namely isolated word recognition. The dynamic nature of speech is considered by adding delayed connection and integration units to the network. We refer to a specific model where the layers are organised in a hirerchical manner: a first RBF hidden layer, a second sigmoidal layer and a classification layer which integrates over time the partial classifications performed by the sigmoidal layer. The training procedures for RBF networks are compared on both generalisation ability and computational costs. Our study shows that supervised learning of the centroids of the basis functions gives appreciable results at a significantly lower cost.",
keywords = "Backpropagation, Isolated word recognition, Multilayer perceptrons, Neural networks, Radial basis functions, Speech processing",
author = "Michele Ceccarelli and Hounsou, {Jo{\"e}l T.}",
year = "1996",
month = "5",
day = "1",
doi = "10.1016/0925-2312(94)00079-4",
language = "English",
volume = "11",
pages = "75--88",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "1",

}

TY - JOUR

T1 - Sequence recognition with radial basis function networks

T2 - Experiments with spoken digits

AU - Ceccarelli, Michele

AU - Hounsou, Joël T.

PY - 1996/5/1

Y1 - 1996/5/1

N2 - In this paper we consider several learning procedures for Radial Basis Function (RBF) Networks applied to a problem of speech recognition, namely isolated word recognition. The dynamic nature of speech is considered by adding delayed connection and integration units to the network. We refer to a specific model where the layers are organised in a hirerchical manner: a first RBF hidden layer, a second sigmoidal layer and a classification layer which integrates over time the partial classifications performed by the sigmoidal layer. The training procedures for RBF networks are compared on both generalisation ability and computational costs. Our study shows that supervised learning of the centroids of the basis functions gives appreciable results at a significantly lower cost.

AB - In this paper we consider several learning procedures for Radial Basis Function (RBF) Networks applied to a problem of speech recognition, namely isolated word recognition. The dynamic nature of speech is considered by adding delayed connection and integration units to the network. We refer to a specific model where the layers are organised in a hirerchical manner: a first RBF hidden layer, a second sigmoidal layer and a classification layer which integrates over time the partial classifications performed by the sigmoidal layer. The training procedures for RBF networks are compared on both generalisation ability and computational costs. Our study shows that supervised learning of the centroids of the basis functions gives appreciable results at a significantly lower cost.

KW - Backpropagation

KW - Isolated word recognition

KW - Multilayer perceptrons

KW - Neural networks

KW - Radial basis functions

KW - Speech processing

UR - http://www.scopus.com/inward/record.url?scp=0030152349&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030152349&partnerID=8YFLogxK

U2 - 10.1016/0925-2312(94)00079-4

DO - 10.1016/0925-2312(94)00079-4

M3 - Article

AN - SCOPUS:0030152349

VL - 11

SP - 75

EP - 88

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 1

ER -