Light Weight Mobile Device Targeted Speaker Clustering Algorithm

Olli Vuorinen, Tommi Lahti, Satu-Marja Mäkelä, Johannes Peltola

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

2 Citations (Scopus)

Abstract

In this paper we present a novel light weight speaker clustering algorithm based on the Bayesian Information Criterion (BIC). Algorithm utilises BIC profiles, which were earlier used for False Alarm Compensation (FAC) in our Speaker Change Detector (SCD). Proposed speaker segmentation followed by a light weight clustering is targeted to segment and label mobile device recordings directly in the device itself. Thus the main criterion in algorithm design was to maintain high detection accuracy while keeping computational costs in low level. Clustering algorithm gave F-score performance of 0.90 for speaker segmentation, which is 29% relative improvement compared to baseline [1] results. Speaker segment labelling performance was 88%, when the number of speakers was undetermined. The experimental results indicate that our unsupervised speaker clustering algorithm is sufficiently effective and efficient for speaker segmentation applications in mobile devices.
Original languageEnglish
Title of host publication2008 IEEE 10th Workshop on Multimedia Signal Processing
Place of PublicationAustralia
PublisherInstitute of Electrical and Electronic Engineers IEEE
Pages131-136
ISBN (Electronic)978-1-4244-2295-1
ISBN (Print)978-1-4244-2294-4
DOIs
Publication statusPublished - 2008
MoE publication typeA4 Article in a conference publication
Event2008 International Workshop on Multimedia Signal Processing, MMSP 2008 - Cairns, Australia
Duration: 8 Oct 200810 Oct 2008

Conference

Conference2008 International Workshop on Multimedia Signal Processing, MMSP 2008
Abbreviated titleMMSP 2008
CountryAustralia
CityCairns
Period8/10/0810/10/08

Fingerprint

Clustering algorithms
Mobile devices
Labeling
Labels
Detectors
Costs

Keywords

  • robustness
  • Gaussian distribution
  • mobiel handsets
  • speech
  • clustering algorithms

Cite this

Vuorinen, O., Lahti, T., Mäkelä, S-M., & Peltola, J. (2008). Light Weight Mobile Device Targeted Speaker Clustering Algorithm. In 2008 IEEE 10th Workshop on Multimedia Signal Processing (pp. 131-136). Australia: Institute of Electrical and Electronic Engineers IEEE. https://doi.org/10.1109/MMSP.2008.4665062
Vuorinen, Olli ; Lahti, Tommi ; Mäkelä, Satu-Marja ; Peltola, Johannes. / Light Weight Mobile Device Targeted Speaker Clustering Algorithm. 2008 IEEE 10th Workshop on Multimedia Signal Processing. Australia : Institute of Electrical and Electronic Engineers IEEE, 2008. pp. 131-136
@inproceedings{71b325de543843d0beca329f2dcf6e0f,
title = "Light Weight Mobile Device Targeted Speaker Clustering Algorithm",
abstract = "In this paper we present a novel light weight speaker clustering algorithm based on the Bayesian Information Criterion (BIC). Algorithm utilises BIC profiles, which were earlier used for False Alarm Compensation (FAC) in our Speaker Change Detector (SCD). Proposed speaker segmentation followed by a light weight clustering is targeted to segment and label mobile device recordings directly in the device itself. Thus the main criterion in algorithm design was to maintain high detection accuracy while keeping computational costs in low level. Clustering algorithm gave F-score performance of 0.90 for speaker segmentation, which is 29{\%} relative improvement compared to baseline [1] results. Speaker segment labelling performance was 88{\%}, when the number of speakers was undetermined. The experimental results indicate that our unsupervised speaker clustering algorithm is sufficiently effective and efficient for speaker segmentation applications in mobile devices.",
keywords = "robustness, Gaussian distribution, mobiel handsets, speech, clustering algorithms",
author = "Olli Vuorinen and Tommi Lahti and Satu-Marja M{\"a}kel{\"a} and Johannes Peltola",
year = "2008",
doi = "10.1109/MMSP.2008.4665062",
language = "English",
isbn = "978-1-4244-2294-4",
pages = "131--136",
booktitle = "2008 IEEE 10th Workshop on Multimedia Signal Processing",
publisher = "Institute of Electrical and Electronic Engineers IEEE",
address = "United States",

}

Vuorinen, O, Lahti, T, Mäkelä, S-M & Peltola, J 2008, Light Weight Mobile Device Targeted Speaker Clustering Algorithm. in 2008 IEEE 10th Workshop on Multimedia Signal Processing. Institute of Electrical and Electronic Engineers IEEE, Australia, pp. 131-136, 2008 International Workshop on Multimedia Signal Processing, MMSP 2008, Cairns, Australia, 8/10/08. https://doi.org/10.1109/MMSP.2008.4665062

Light Weight Mobile Device Targeted Speaker Clustering Algorithm. / Vuorinen, Olli; Lahti, Tommi; Mäkelä, Satu-Marja; Peltola, Johannes.

2008 IEEE 10th Workshop on Multimedia Signal Processing. Australia : Institute of Electrical and Electronic Engineers IEEE, 2008. p. 131-136.

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Light Weight Mobile Device Targeted Speaker Clustering Algorithm

AU - Vuorinen, Olli

AU - Lahti, Tommi

AU - Mäkelä, Satu-Marja

AU - Peltola, Johannes

PY - 2008

Y1 - 2008

N2 - In this paper we present a novel light weight speaker clustering algorithm based on the Bayesian Information Criterion (BIC). Algorithm utilises BIC profiles, which were earlier used for False Alarm Compensation (FAC) in our Speaker Change Detector (SCD). Proposed speaker segmentation followed by a light weight clustering is targeted to segment and label mobile device recordings directly in the device itself. Thus the main criterion in algorithm design was to maintain high detection accuracy while keeping computational costs in low level. Clustering algorithm gave F-score performance of 0.90 for speaker segmentation, which is 29% relative improvement compared to baseline [1] results. Speaker segment labelling performance was 88%, when the number of speakers was undetermined. The experimental results indicate that our unsupervised speaker clustering algorithm is sufficiently effective and efficient for speaker segmentation applications in mobile devices.

AB - In this paper we present a novel light weight speaker clustering algorithm based on the Bayesian Information Criterion (BIC). Algorithm utilises BIC profiles, which were earlier used for False Alarm Compensation (FAC) in our Speaker Change Detector (SCD). Proposed speaker segmentation followed by a light weight clustering is targeted to segment and label mobile device recordings directly in the device itself. Thus the main criterion in algorithm design was to maintain high detection accuracy while keeping computational costs in low level. Clustering algorithm gave F-score performance of 0.90 for speaker segmentation, which is 29% relative improvement compared to baseline [1] results. Speaker segment labelling performance was 88%, when the number of speakers was undetermined. The experimental results indicate that our unsupervised speaker clustering algorithm is sufficiently effective and efficient for speaker segmentation applications in mobile devices.

KW - robustness

KW - Gaussian distribution

KW - mobiel handsets

KW - speech

KW - clustering algorithms

U2 - 10.1109/MMSP.2008.4665062

DO - 10.1109/MMSP.2008.4665062

M3 - Conference article in proceedings

SN - 978-1-4244-2294-4

SP - 131

EP - 136

BT - 2008 IEEE 10th Workshop on Multimedia Signal Processing

PB - Institute of Electrical and Electronic Engineers IEEE

CY - Australia

ER -

Vuorinen O, Lahti T, Mäkelä S-M, Peltola J. Light Weight Mobile Device Targeted Speaker Clustering Algorithm. In 2008 IEEE 10th Workshop on Multimedia Signal Processing. Australia: Institute of Electrical and Electronic Engineers IEEE. 2008. p. 131-136 https://doi.org/10.1109/MMSP.2008.4665062