Speaker verification in score-ageing-quality classification space

Kelly, Finnian; Drygajlo, Andrzej; Harte, Naomi

doi:10.1016/j.csl.2012.12.005

Cited by 30 publications

(16 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Recent research on use of GMM for speaker verification [14] confirms the validity of the experiments carried out in this paper. Further investigation is needed to evaluate the validity of the technique provided here on real-word speaker verification tasks.…”

Section: Resultssupporting

confidence: 54%

Speaker model adaptation based on confidence score

Mengusoglu

2015

Teh. vjesn.

View full text Add to dashboard Cite

Original scientific paper Confidence measures are expected to give a measure of reliability on the result of a speech/speaker recognition system. Most commonly used confidence measures are based on posterior word or phoneme probabilities which can be obtained from the output of the recognizer. In this paper we introduced a linear interpretation of posterior probability based confidence measure by using inverse Fisher transformation. Speaker adaptation consists in updating model parameters of a speaker independent model to have a better representation of the current speaker. Confidence measures give more reliable selection criteria to select the utterances which best represent the speaker. A linear interpretation of confidence measure is very important to select the most representative data for adaptation. Keywords: confidence measure; Fisher transformation; speaker adaptation; speaker verification Adaptacija modela govornika na osnovu rezultata povjerenjaIzvorni znanstveni članak Očekuje se da mjere povjerenja postanu mjera za pouzdanost rezultata sustava za prepoznavanje govora. Najčešće korištene mjere povjerenja zasnovane su na vjerojatnosti sljedeće riječi ili fonema, koja se može dobiti iz izlaznog rezultata prepoznavatelja. U ovom smo radu uveli mjeru povjerenja zasnovanu na linearnoj interpretaciji vjerojatnoće sljedeće riječi primjenom obrnute Fisher transformacije. Adaptacija govornika sastoji se od ažuriranja parametara modela nezavisnog od govornika zbog boljeg predstavljanja postojećeg govornika. Mjere povjerenja daju pouzdanije kriterije za odabir riječi koje najbolje predstavljaju govornika. Linearna interpretacija mjere povjerenja vrlo je važna pri odabiru najreprezentativnijih podataka za adaptaciju.

show abstract

Section: Resultssupporting

confidence: 54%

Speaker model adaptation based on confidence score

Mengusoglu

2015

Teh. vjesn.

View full text Add to dashboard Cite

show abstract

“…o Between-session analysis: will evaluate the effect of incorporating longitudinal speaker data in PLDA training. Long-term longitudinal data from TCDSA [Kelly13] will be used for this purpose. We will also explore lightly supervised data harvesting from YouTube, where audio from bloggers, or other users with regular uploads across time, will be identified, downloaded and then processed automatically.…”

Section: Multi-session Sid Ongoing/future Workmentioning

confidence: 99%

Robust Speech Processing & Recognition: Speaker ID, Language ID, Speech Recognition/Keyword Spotting, Diarization/Co-Channel/Environmental Characterization, Speaker State Assessment

Hansen¹

2015

View full text Add to dashboard Cite

Using Government drawings, specifications, or other data included in this document for any purpose other than Government procurement does not in any way obligate the U.S. Government. The fact that the Government formulated or supplied the drawings, specifications, or other data does not license the holder or any other person or corporation; or convey any rights or permission to manufacture, use, or sell any patented invention that may relate to them. This report is the result of contracted fundamental research deemed exempt from public affairs security and policy review in accordance with SAF/AQR memorandum dated 10 Dec 08 and AFRL/CA policy clarification memorandum dated 16 Jan 09. This report is available to the general public, including foreign nationals. Copies may be obtained from the Defense Technical Information Center (DTIC) (http://www.dtic.mil).

show abstract

“…To our knowledge, databases of voice samples exist for aging voices (Kelly, Drygajlo, & Harte, 2013), pathological speech/voices (e.g., Disordered Voice Database, Kay Elemetrics Corp., Lincoln Park, NJ, USA), and emotional voices (Belin, Fillion-Bilodeau, & Gosselin, 2008;Burkhardt, Paeschke, Rolfes, Sendlmeier, & Weiss, 2005;Petta, Pelachaud, & Cowie, 2011). Most facial databases were designed for the study of face recognition (Gross, 2005;www. face-rec.org/databases) and/or facial expression recognition, such as the Pictures of Facial Affect (Ekman & Friesen, 1976), the Karolinska Directed Emotional Faces (Lundqvist, Flykt, & Öhman, 1998), the PAL (Minear & Park, 2004), the Yale Face Database (Belhumeur, Hespanha, & Kriegman, 1997), the PICS (University of Stirling, http://pics.psych.stir.ac.uk), and the FACES (Max Planck Institute for Human Development in Berlin, http://faces.mpib-berlin.mpg.de).…”

mentioning

confidence: 99%

The Geneva Faces and Voices (GEFAV) database

et al. 2014

View full text Add to dashboard Cite

Although many facial and vocal databases are available for research, very few of them have controlled the range of attractiveness of the stimuli that they offer. To fill this gap, we created the GEneva Faces and Voices (GEFAV) database, providing standardized faces (static and dynamic neutral, smiling) and voices (speaking sentences, vowels) of young European adults. A total of 61 women and 50 men 18-35 years old agreed to be part of the GEFAV stimuli, and two rating studies involving 285 participants provided evaluations of the facial and vocal samples. The final set of stimuli was satisfactory in terms of attractiveness range (wide and rather symmetrical distribution over the attractiveness continuum) and the reliability of the ratings (high consistency between the two rating studies, high interrater agreement in the final rating study). Moreover, the database showed an adequate validity, since a series of findings described by earlier research on human attractiveness were confirmed-namely, that facial and vocal attractiveness are predicted by femininity and health in women, and by masculinity, dominance, and trustworthiness in men. In future studies, the GEFAV stimuli may be used intact or transformed, individually or in multimodal combinations, to investigate a wide range of mechanisms, such as the behavioral, neuropsychological, and neurophysiological processes involved in social cognition.

show abstract

Speaker verification in score-ageing-quality classification space

Cited by 30 publications

References 42 publications

Speaker model adaptation based on confidence score

Speaker model adaptation based on confidence score

Robust Speech Processing & Recognition: Speaker ID, Language ID, Speech Recognition/Keyword Spotting, Diarization/Co-Channel/Environmental Characterization, Speaker State Assessment

The Geneva Faces and Voices (GEFAV) database

Contact Info

Product

Resources

About