In-between speech: the speaker-specificity of non-speech vocalisations

During a crime, victims or witnesses may hear a perpetrator's voice without seeing their face and
are later asked to identify them from a voice identification line-up. Elsewhere, an analyst compares
an incriminating recording of a perpetrator with a sample of a suspect's speech. The field of
forensic speech science is concerned with how to identify a person from their voice, but sometimes
the two samples being compared are mismatched. In the situation of a crime, there may not be
much actual speech - instead, there may be only screams, groans, or laughter, known as non-speech vocalisations (NSVs). In the compared sample or voice line-up, however, there is typically
only speech, and NSVs are discarded. What is currently unknown is whether speakers can
accurately be identified from NSVs. This project will investigate this question by asking how
accurately listeners identify speakers from different NSV types, and how speaker-specific acoustic
features vary in different NSV types. It will also explore whether speakers who are easily recognised
by their NSVs are also equally easily recognised by their speech. This will benefit the field by
improving our understanding of how speakers can be identified by NSVs.