The Influence of Signal Complexity on Speaker Identification

Abstract

Multiple factors may affect a listener’s ability to identify a particular voice. For example, speaker eff ects such as language and dialect have been shown to negatively impact speaker identifi cation accuracy (Bahr and Frisch 2002, Doty 1998). Mechanical eff ects such as communication channel (i.e. cell phone, land line) have also been shown to negatively impact voice identifi cation (Künzel 2001). However, the contribution of these factors to signal complexity as a whole has not been well-defi ned. Therefore, it is the purpose of this investigation to assess the effect of signal complexity on speaker identification. A paired comparison listening task was used to evaluate monolingual listeners’ performance when channel (lab quality, landline phone and cell phone transmissions), language (English vs. Spanish), and dialect (different dialects of Spanish) were varied. Accuracy, reaction time, and d-prime scores were use to measure the effect of signal complexity on speaker identifi cation accuracy. Results indicated that listeners were less accurate when the signal was most complex. These findings were across all three measures. The role of signal complexity in the forensic speaker identification process will be described.