Filling the population statistics gap: Swiss German reference data on F0 and speech tempo for forensic contexts

The increased focus on big data in phonetics, and Bayesian statistics in the forensic sciences, prompt a fundamental issue in common applications of forensic phonetics. Relevant population distributions for most features, a key element when evaluating the similarity and distinctiveness of voices, remain lacking for a substantial number of languages and dialects. This paper provides population statistics for two phonetic features in the Swiss German context, speech tempo and F0, and outlines a potential method for big data analysis. The speech data is taken from 1000 SwG speakers and include two different style conditions: spontaneous and read speech. Results indicate significant variation for both parameters: we contradict previous findings on gender differences in speech tempo and note discrepancies for both features between the two styles. These findings constitute an important contribution to the field of forensic phonetics, as well as the field of general phonetics more broadly.