A COMPARISON OF INDICES OF DIFFERENCE AND SIMILARITY, BASED ON LTASS AND TESTED ON VOICES IN REAL FORENSIC CASE AND IN CONTROLLED CONDITIONS

Gordana Varošanec-Škaric & Jordan Bicanic
Faculty of Humanities and Social Sciences, Dpt. of Phonetics, Univeritiy of Zagreb

ID 1047
[full paper]

For 43 pairs of same and different voices in different speech contexts of real cases, LTASS has been made for the area between 800 and 3500 Hz on the basis of the recordings of speech over GSM mobile phones. SDDD and similarity indices in different speech context have been compared. For the sake of comparison of data achieved by the two speech recordings of standardized text read by 30 male and 35 female speakers, average values of indices are calculated for the same and different people: from 0 to 10 kHz and filtered voices from 0.8 to 3.5 kHz. The results of t-test have shown that the groups differ significantly, respectively greatest in the group of male voices recorded in the studio (0-10 kHz: p<0.001), filtered studio voices (p<0.001) and real cases (SDDD: p<0.01; R: p=0.01).