The performance of human listeners is the final objective of the automated speech input systems. In speaker recognition the need of a human-reference to assess automated method is a common procedure. We address the problem of identify unfamiliar voices, assuming that the listeners use a limited amount of speech to create the reference template. Our aim is to define and assess standard procedures to evaluate listener's capability in speaker recognition.
Aural Identification of Unfamiliar Voices by Comparison
Contributo in atti di convegno
XIII International Congress of Phonetic Sciences ICPhS-95, pp. 186–189, Stochkolm, Sweden, 14-18 August, 1995