Combining multiple estimators of speaking rate

Published 27 November 2002

N. Morgan, Eric Fosler‐Lussier

DOI10.1109/icassp.1998.675368

Citations127

TL;DR

The development of a measure of speaking rate that is computed from the acoustic signal, called mrate, is reported, and the use of mrate as a detector for rapid speech is assessed.

Abstract

We report progress in the development of a measure of speaking rate that is computed from the acoustic signal. The newest form of our analysis incorporates multiple estimates of rate; besides the spectral moment for a full-band energy envelope that we have previously reported, we also used pointwise correlation between pairs of compressed sub-band energy envelopes. The complete measure, called mrate, has been compared to a reference syllable rate derived from a manually transcribed subset of the Switchboard database. The correlation with transcribed syllable rate is significantly higher than our earlier measure; estimates are typically within 1-2 syllables/second of the reference syllable rate. We conclude by assessing the use of mrate as a detector for rapid speech.

Keywords

PsychologyComputer Science

The Journal of the Acoustical Society of AmericaParallel Processing Techniques for Estimating Pitch Periods of Speech in the Time Domain

241 Citations1969Ben Gold, L. R. Rabiner

A computational algorithm for estimating pitch periods of speech in the time domain is presented, and two recent modifications of the algorithm are discussed in detail.

On the effects of speech rate in large vocabulary speech recognition systems

92 Citations2002M.A. Siegler, Richard M. Stern

It is suggested that phone rate is a more meaningful measure of speech rate than the more common word rate, and it is found that when data sets are clustered according to the phone rate metric, recognition errors increase when thePhone rate is more than 1 standard deviation greater than the mean.

Speech recognition using on-line estimation of speaking rate

87 Citations1997Nelson Morgan, Eric Fosler +1 more

It is shown that a simple measure of sp eaking rate that is directly based onSignal pro cessing of the sp eech, without referencetolexical units or requiring the use of a recognition sys-temforestimation can reduce the error rate almost as much as using lexical phones p er second calculated from man-ually transcrib ed data.

Towards robustness to fast speech in ASR

46 Citations2002Nikki Mirghafori, Eric Fosler +1 more

This work has studied various measures for quantifying rate of speech (ROS) and used simple methods for estimating the speaking rate of a novel utterance using ASR technology and implemented mechanisms that make the ASR system more robust to fast speech.

A fast and reliable rate of speech detector

38 Citations2002Jan Verhasselt, J.-P. Martens

A new rate-of-speech (ROS) detector that operates independently from the recognition process is presented and the ROS estimate is subsequently used to compensate for the effects of unusual speech rates on continuous speech recognition.

Extraction and representation rhythmic components of spontaneous speech

12 Citations1997Shigeyoshi Kitaazawa, Hideya Ichikawa +2 more

The TEMAX-gram, a sonagraphic output of speech envelope, the DFT using a 1-second window is convenient to set off isosyllabic characteristics, and the periodicity of syllables between stress is displayed as RF2.