Final published version
Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review
Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review
}
TY - GEN
T1 - Cross-linguistic speaker individuality of long-term formant distributions: Phonetic and forensic perspectives
AU - Lo, Justin J. H.
PY - 2021/9/1
Y1 - 2021/9/1
N2 - This study considers issues of language- and speakerspecificity in long-term formant distributions (LTFDs) from phonetic and forensic perspectives and examines their potential value in cases of cross-language forensic voice comparison. Acoustic analysis of 60 male English-French bilinguals revealed systematic differences in LTFDs between the two languages, with higher LTF2-4 in French than in English. Cross-linguistic differences in the shapes of LTFDs were also found. These differences are argued to reflect not only vowel inventories of each language but also languagespecific phonetic settings. At the same time, a high degree of within-speaker consistency was found across languages. Likelihood ratio based testing was carried out to examine the effect of language mismatch on the utility of LTFDs as speaker discriminants. Results showed that while the performance of LTFDs was worse in cross-language comparisons than in same-language comparisons, they were still capable of providing speaker-specific information. These findings demonstrate that, in spite of deteriorated performance, LTFDs are still potentially useful speaker discriminants in cases of language mismatch. These findings thus call for further empirical investigation into the use of linguistic-phonetic features in cross-language comparisons.
AB - This study considers issues of language- and speakerspecificity in long-term formant distributions (LTFDs) from phonetic and forensic perspectives and examines their potential value in cases of cross-language forensic voice comparison. Acoustic analysis of 60 male English-French bilinguals revealed systematic differences in LTFDs between the two languages, with higher LTF2-4 in French than in English. Cross-linguistic differences in the shapes of LTFDs were also found. These differences are argued to reflect not only vowel inventories of each language but also languagespecific phonetic settings. At the same time, a high degree of within-speaker consistency was found across languages. Likelihood ratio based testing was carried out to examine the effect of language mismatch on the utility of LTFDs as speaker discriminants. Results showed that while the performance of LTFDs was worse in cross-language comparisons than in same-language comparisons, they were still capable of providing speaker-specific information. These findings demonstrate that, in spite of deteriorated performance, LTFDs are still potentially useful speaker discriminants in cases of language mismatch. These findings thus call for further empirical investigation into the use of linguistic-phonetic features in cross-language comparisons.
KW - Bilingual
KW - Forensic voice comparison
KW - Long-term formant distributions
KW - Speaker specificity
U2 - 10.21437/Interspeech.2021-1699
DO - 10.21437/Interspeech.2021-1699
M3 - Conference contribution/Paper
T3 - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
SP - 4615
EP - 4619
BT - 22nd Annual Conference of the International Speech Communication Association, INTERSPEECH 2021
ER -