Universiteit Leiden

nl en


LACG: Willemijn Heeren

  • Willemijn Heeren
5 April 2018
Language and Cognition Group meetings
Matthias de Vrieshof
Matthias de Vrieshof 1
2311 BZ Leiden

The speaker in speech – background and (very) preliminary results

The NWO VIDI project The Speaker in Speech (2017-2022) investigates the relationship between linguistic and indexical information in speech to discover how speaker-specific information is distributed across the speech signal. In experimental phonetic and psycholinguistic research individual speakers’ behavior has traditionally been disregarded. For instance, differences between speakers’ voices were thought to be normalized away from the auditory input during word recognition and experimental results were typically presented as means across participants. Only recently, and conveniently supported by techniques such as mixed-effects modeling, individual language users have become much more important, in both production and perception research. In this talk I will first present the background of this research project in forensic phonetics. The general question in the field is which acoustic-phonetic features in speech are speaker-dependent, to support forensic voice comparisons. The central questions of the research project are how and why the speaker-specificity of speech sounds depends on different aspects of the spoken message: the speech sound’s linguistic context (e.g. in a stressed syllable or not), the speech channel (e.g. over the telephone), and the language spoken. I will further present the first stage of preparatory work that has yielded enriched speech databases for potential use by others, and the project’s initial experimental results.