Speech utterance clustering based on the maximization of within-cluster homogeneity of speaker voice characteristics

Tsai, Wei-Ho; Wang, Hsin‐Min

doi:10.1121/1.2225570

Cited by 6 publications

(2 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In any event, the importance of research on speech features should not be underestimated. For example, many different (internally related) clusters of these parameters can be most useful in building speech vectors . An especially effective approach to SI is when the multiparameter vectors are, in turn, combined into a multivector profile of the speaker .…”

Section: Part Ii: Speech Analysis: An Illustrative Approach To Simentioning

confidence: 99%

An Approach to Speaker Identification

Hollien

2016

Journal of Forensic Sciences

View full text Add to dashboard Cite

This presentation will provide standards upon which any attempts to meet the challenge of identifying speakers by voice should be based. It is organized into a model based on (i) application of a rigorous research program validating the system, (ii) an upgrading of the organization of the SI area, and (iii) exploitation of new technology. The second part of the presentation will describe an illustrative speech/voice approach to SI development. This effort is also based on an extensive corpus of research. It is suggested that application of the cited standards, plus the illustrative model, will permit reasonable progress to be made. Finally, a number of procedural recommendations are made; they should enhance the efficacy of the proposed approach.

show abstract

Section: Part Ii: Speech Analysis: An Illustrative Approach To Simentioning

confidence: 99%

An Approach to Speaker Identification

Hollien

2016

Journal of Forensic Sciences

View full text Add to dashboard Cite

show abstract

“…In turn, this trend has led to the development of a number of automatic speaker verification systems. For example, some of the early ones showed promise 51 , as do several of the more recent ones 5,50,59 . Unfortunately, however, very little confirmatory data -hard data, that is -about either their success or failure is available.…”

Section: The Engineersmentioning

confidence: 99%

Barriers to Progress in Speaker Identification with Comments on the Trayvon Martin Case

Hollien

2013

LESLI

View full text Add to dashboard Cite

Linguistics and phonetics overlap in many areas. The essay to follow reviews some of the problems experienced by phoneticians in one of these regions. It may provide some insight for linguists when they are confronted by barriers in their own field. The present example involves individuals who are attempting to identify speakers from voice analysis. The fundamental challenge they face is, of course, caused by the thousands of variables associated with that task. Included here are differences among speakers' gender, age, size, physiology, language, dialect, psychological/health states, background/education, reason for speaking, situation, environment, configuration of the acoustic channel --plus many others. Many formal assessment procedures --both aural-perceptual ones conducted by humans or machine/computer based systems --have been proposed and/or used for the cited analyses. Unfortunately, however, few have enjoyed particularly high levels of success. Worse yet, reasonable progress has suffered from external impedances; the report to follow will outline some of them. Among the problems considered are: 1) competition (verification vs. identification, from voiceprints), 2) concept disputes 3) the continued undervaluation of relevant evidence and 4) markedly dissimilar philosophies of professionals from different disciplines. A response in the form of a short review of the data and concepts which clearly support the possibility of robust speaker identification is presented. Also included are suggestions as to how to enhance the effectiveness of disciplines such as ours.

show abstract

Evolutionary minimization of the Rand index for speaker clustering

Tsai

Wang

2009

Computer Speech & Language

Self Cite

View full text Add to dashboard Cite

We propose an effective method for clustering unknown speech utterances based on their associated speakers. The method jointly optimizes the generated clusters and the required number of clusters by estimating and minimizing the Rand index. The metric reflects the clustering errors that arise when utterances from the same speaker are placed in different clusters; or when utterances from different speakers are placed in the same cluster. One useful characteristic of the Rand index is that its value only reaches the minimum when the number of clusters is equal to the size of the true speaker population. We approximate the Rand index by a function of the similarity measures between utterances and then use a genetic algorithm to determine the cluster in which each utterance should be located, such that the function is minimized. Our experiment results show that this novel speaker-clustering method outperforms conventional methods that use the Bayesian information criterion to determine the required number of clusters.

show abstract

Speech utterance clustering based on the maximization of within-cluster homogeneity of speaker voice characteristics

Cited by 6 publications

References 23 publications

An Approach to Speaker Identification

An Approach to Speaker Identification

Barriers to Progress in Speaker Identification with Comments on the Trayvon Martin Case

Evolutionary minimization of the Rand index for speaker clustering

Contact Info

Product

Resources

About