Examining long-term formant distributions as a discriminant in forensic speaker comparisons under a likelihood ratio framework

Gold, Erica; French, Peter; Harrison, Philip

doi:10.1121/1.4800285

Cited by 11 publications

(17 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Goldova i saradnici (2013) istraživali su dugoročne frekvencije prva četiri formanta kod 100 izvornih govornika engleskog uz pomoć Bajesove formule verovatnoće. Prema rezultatima tog istraživanja, DFF3 se pokazao kao najbolji diskriminant sa najmanjim stepenom greške, a najbolji rezultati su dobijeni kombinacijom sva četiri dugoročna formanta (Gold et al 2013).…”

Section: Prethodna Istraživanja Dugoročnih Frekvencija Formanataunclassified

“…Nakon obrade, dobili smo uzorke u prosečnom trajalju od 46,8 s (SD 8,24) za govor na srpskom i 46,2 s (SD 8,33) za govor na engleskom jeziku. Frekvencije formanata su izmerene na 10 milisekundi uz pomoć Burgovog algoritma baziranog na linearnoj predikciji, zatim smo svakih sto mera udružili u takozvane "pakete" i izračunali srednju vrednost svakog "paketa", te je u nastavku ta srednja vrednost posmatrana kao jedno merenje (v. Gold et al 2013). Na ovaj način, dobili smo vrednosti takozvanih "lokalnih" dugoročnih frekvencija formanata.…”

Section: Akustička I Statistička Analizaunclassified

See 1 more Smart Citation

Međujezična forenzička komparacija glasova — dugoročne frekvencije formanata

Tomić¹

2020

Примењена лингвистика

View full text Add to dashboard Cite

Filozofski fakultet u NišuMEĐUJEZIČNA FORENZIČKA KOMPARACIJA GLASOVA -DUGOROČNE FREKVENCIJE FORMANATA Sažetak: Međujezična forenzička komparacija glasova (FKG) je jedan od velikih izazova u oblasti forenzičke fonetike. Cilj ovog istraživanja je da se utvrdi stabilnost dugoročnih frekvencija formanata (DFF) kao parametra u FKG kada je sporni uzorak na stranom a nesporni na maternjem jeziku. Deset izvornih govornica srpskog snimljeno je u spontanom govoru na srpskom i engleskom jeziku. Rezultati su pokazali da postoje značajne razlike u DFF između govornica kako u govoru na maternjem tako i u govoru na stranom jeziku. Dugoročni F4 i F3 se ističu kao dobri diskriminanti jer je za njih interspikerska varijacija značajno veća od intraspikerske varijacije u oba jezika.

show abstract

Section: Prethodna Istraživanja Dugoročnih Frekvencija Formanataunclassified

Section: Akustička I Statistička Analizaunclassified

Međujezična forenzička komparacija glasova — dugoročne frekvencije formanata

Tomić¹

2020

Примењена лингвистика

View full text Add to dashboard Cite

show abstract

“…This may be due to the fact that we used the same default settings for formant extraction in the Snack Toolkit [17], with an LPC order of 12 and tracking four formants, for all speakers across all conditions. Although choosing system-level settings is the approach that has been followed in previous studies [5,6,7], it may have led to measurement issues here. Speakers with inherently high F3 are more likely to have F4 values close to or outside the upper bandpass threshold for telephone transmission.…”

Section: Individualsmentioning

confidence: 99%

“…A growing body of research has shown that SASR based on LTFDs can be of considerable value. [5] reported an equal error rate (EER) of 4.14% based on F1, F2, F3, and F4 frequency values extracted from contemporaneous high quality studio recordings of British English speakers. [6] found comparable results for high quality studio recordings of German, with EERs ranging from 3% to 10.5%.…”

Section: ! Introductionmentioning

confidence: 99%

The Individual and the System: Assessing the Stability of the Output of a Semi-automatic Forensic Voice Comparison System

Hughes¹,

Harrison²,

Foulkes³

et al. 2018

Interspeech 2018

Self Cite

View full text Add to dashboard Cite

Semi-automatic systems based on traditional linguisticphonetic features are increasingly being used for forensic voice comparison (FVC) casework. In this paper, we examine the stability of the output of a semi-automatic system, based on the long-term formant distributions (LTFDs) of F1, F2, and F3, as the channel quality of the input recordings decreases. Crossvalidated, calibrated GMM-UBM log likelihood-ratios (LLRs) were computed for 97 Standard Southern British English speakers under four conditions. In each condition the same speech material was used, but the technical properties of the recordings changed (high quality studio recording, landline telephone recording, high bit-rate GSM mobile telephone recording and low bit-rate GSM mobile telephone recording). Equal error rate (EER) and the log LR cost function (C llr) were compared across conditions. System validity was found to decrease with poorer technical quality, with the largest differences in EER (21.66%) and C llr (0.46) found between the studio and the low bit-rate GSM conditions. However, importantly, performance for individual speakers was affected differently by channel quality. Speakers that produced stronger evidence overall were found to be more variable. Mean F3 was also found to be a predictor of LLR variability, however no effects were found based on speakersÕ voice quality profiles.

show abstract

“…The linear mixed effects and ANOVA results illustrated that F1 and F2 are more regionally influenced than F3. One explanation for why F3 may vary the least across boroughs is that this parameter is generally considered to be more idiosyncratic [15,16,17] and therefore less likely to be dependent on external factors, such as the place the speaker is from.…”

Section: General Findingsmentioning

confidence: 99%

Variation in the FACE Vowel across West Yorkshire: Implications for Forensic Speaker Comparisons

Earnshaw

Gold

2018

Interspeech 2018

Self Cite

View full text Add to dashboard Cite

In the field of sociophonetics, research is largely focused on the documentation of regional variability. However, the majority of literature in the United Kingdom often reports on variation at a macro-level (e.g. Northern, Yorkshire, West Yorkshire) rather than at a more local level (e.g. West Yorkshire: Bradford, Calderdale, Kirklees, Leeds, Wakefield). Traditionally, for sociophoneticians, examining regional variation at a broader level is adequate for answering research questions related to language change or more general variation. For practical applications (e.g. forensic, speech technology), however, more fine-grained regional analysis is necessary. This paper analyses over 2000 FACE tokens from three metropolitan boroughs (Bradford, Kirklees and Wakefield) within West Yorkshire, in order to determine the extent to which F1~F3 vary across the region. Results suggest that for FACE, these three boroughs within West Yorkshire are more regionally stratified than previously acknowledged. These findings are of particular importance to the forensic speech science community, as experts rely on these regional nuances in order to make important judgments related to strength of the speech evidence in a case. Should decisions be made without the greater understanding of local-level variation, the strength of evidence risks being over-or underestimated .

show abstract

Examining long-term formant distributions as a discriminant in forensic speaker comparisons under a likelihood ratio framework

Cited by 11 publications

References 0 publications

Međujezična forenzička komparacija glasova — dugoročne frekvencije formanata

Međujezična forenzička komparacija glasova — dugoročne frekvencije formanata

The Individual and the System: Assessing the Stability of the Output of a Semi-automatic Forensic Voice Comparison System

Variation in the FACE Vowel across West Yorkshire: Implications for Forensic Speaker Comparisons

Contact Info

Product

Resources

About