Automatic speaker recognition as a measurement of voice imitation and conversion

Mireia Farrus, Michael Wagner, Daniel Erro, Javier Hernando

    Research output: Contribution to journalArticle

    11 Citations (Scopus)

    Abstract

    Voices can be deliberately disguised by means of human imitation or voice conversion. The question arises as to what extent they can be modified by using either of both methods. In the current paper, a set of speaker identification experiments are conducted; first, analysing some prosodic features extracted from voices of professional impersonators attempting to mimic a target voice and, second, using both intragender and crossgender converted voices in a spectral-based speaker recognition system. The results obtained in the current experiments show that the identification error rate increases when testing with imitated voices, as well as when using converted voices, especially the crossgender ones.
    Original languageEnglish
    Pages (from-to)119-142
    Number of pages24
    JournalInternational Journal of Speech, Language and the Law
    Volume17
    Issue number1
    DOIs
    Publication statusPublished - 2010

    Fingerprint

    imitation
    experiment

    Cite this

    Farrus, Mireia ; Wagner, Michael ; Erro, Daniel ; Hernando, Javier. / Automatic speaker recognition as a measurement of voice imitation and conversion. In: International Journal of Speech, Language and the Law. 2010 ; Vol. 17, No. 1. pp. 119-142.
    @article{845e4f377a0541aca9b5a237adb45f92,
    title = "Automatic speaker recognition as a measurement of voice imitation and conversion",
    abstract = "Voices can be deliberately disguised by means of human imitation or voice conversion. The question arises as to what extent they can be modified by using either of both methods. In the current paper, a set of speaker identification experiments are conducted; first, analysing some prosodic features extracted from voices of professional impersonators attempting to mimic a target voice and, second, using both intragender and crossgender converted voices in a spectral-based speaker recognition system. The results obtained in the current experiments show that the identification error rate increases when testing with imitated voices, as well as when using converted voices, especially the crossgender ones.",
    author = "Mireia Farrus and Michael Wagner and Daniel Erro and Javier Hernando",
    year = "2010",
    doi = "10.1558/ijsll.v17i1.119",
    language = "English",
    volume = "17",
    pages = "119--142",
    journal = "Forensic Linguistics",
    issn = "1350-1771",
    publisher = "Equinox Publishing Ltd",
    number = "1",

    }

    Automatic speaker recognition as a measurement of voice imitation and conversion. / Farrus, Mireia; Wagner, Michael; Erro, Daniel; Hernando, Javier.

    In: International Journal of Speech, Language and the Law, Vol. 17, No. 1, 2010, p. 119-142.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - Automatic speaker recognition as a measurement of voice imitation and conversion

    AU - Farrus, Mireia

    AU - Wagner, Michael

    AU - Erro, Daniel

    AU - Hernando, Javier

    PY - 2010

    Y1 - 2010

    N2 - Voices can be deliberately disguised by means of human imitation or voice conversion. The question arises as to what extent they can be modified by using either of both methods. In the current paper, a set of speaker identification experiments are conducted; first, analysing some prosodic features extracted from voices of professional impersonators attempting to mimic a target voice and, second, using both intragender and crossgender converted voices in a spectral-based speaker recognition system. The results obtained in the current experiments show that the identification error rate increases when testing with imitated voices, as well as when using converted voices, especially the crossgender ones.

    AB - Voices can be deliberately disguised by means of human imitation or voice conversion. The question arises as to what extent they can be modified by using either of both methods. In the current paper, a set of speaker identification experiments are conducted; first, analysing some prosodic features extracted from voices of professional impersonators attempting to mimic a target voice and, second, using both intragender and crossgender converted voices in a spectral-based speaker recognition system. The results obtained in the current experiments show that the identification error rate increases when testing with imitated voices, as well as when using converted voices, especially the crossgender ones.

    U2 - 10.1558/ijsll.v17i1.119

    DO - 10.1558/ijsll.v17i1.119

    M3 - Article

    VL - 17

    SP - 119

    EP - 142

    JO - Forensic Linguistics

    JF - Forensic Linguistics

    SN - 1350-1771

    IS - 1

    ER -