Talker identification under adverse auditory conditions-The impacts of noise, channel, language, and familiarity

Talker identification is a crucial auditory skill that underpins human social communication and forensic applications. However, real-world conditions pose several challenges-such as environmental noise, channel variability, language familiarity, and talker familiarity-that can undermine the accuracy of auditory identification. In light of the limitations and insights from previous studies, the present study employed auditory experiments to systematically examine the impact of these four adverse factors on talker identification.

The study aimed to address two questions: (1) whether the independent and interactive effects among these factors are significant, and (2) whether lab-training can enhance talker identification accuracy. Using a voice line-up paradigm, this study conducted a perception experiment where speech stimuli were presented under four primary conditions: noise (No Noise vs. Noise), channel (High-quality vs. High-quality; Landline vs. Landline, High-quality vs. Landline), language (Mandarin, Reversed Mandarin, English, Reversed English), and speaker familiarity (assessed through listening tests and lab training). Auditory responses to the stimuli under these adverse conditions were collected from 53 listeners.

The findings indicate that environmental noise and channel variability have significantly negative effects on talker identification, while intelligible speech yields superior performance under adverse conditions compared to unintelligible reversed speech. Furthermore, the study found that lab-training (i.e., increasing talker familiarity) could enhance talker identification accuracy under adverse conditions, although it does not improve accuracy under no noise and high-quality channel conditions.

This paper systematically examines the interactive effects of multiple adverse factors on talker identification, thereby advancing our understanding of the auditory mechanisms underlying human social speech communication and providing important theoretical support for auditory examination techniques in forensic speaker identification.