Speaker discrimination: citation tones vs. coarticulated tones

Abstract

The task of forensic voice comparison (FVC) often involves the comparison of a voice in an offender recording with that in a suspect recording, with the aim to assist the investigating authority or the court in determining the identity of the speaker. One of the main goals in FVC research is to identify speech variables that are useful for differentiating speakers. While French and Stevens (2013) stated that connected speech processes (CSPs) vary across speakers and thus CSPs may be included in the 'toolbox' for forensic voice comparison casework, little empirical research has been done to test how effective various CSPs are in speaker discrimination. This paper reports an exploratory study comparing the speaker-discriminatory power of lexical tones in their citation forms and coarticulated tones. 20 Cantonese and 20 Mandarin speakers were instructed to produce tones under different speech rates and tonal contexts. Results based on discriminant analysis show that the combination of normal speech rate and compatible tonal context appears to have yielded the best speaker discrimination. On the other hand, the combination of fast speech and a conflicting tonal context, which in principle led to the greatest tonal coarticulatory effects, yielded the worst speaker discrimination. The addition of duration on top of tonal f0 significantly improved the classification rates in both languages. Furthermore, for the same tone categories, the Mandarin ones generally discriminate speakers better than the Cantonese counterparts, suggesting that tone inventory density affects the speaker-discriminatory power of tones. Implications of the findings for forensic speaker comparison are discussed.