IAFPA 2006 On sentence content, speaker familiarity and dialect Elisabeth Zetterholm, Erik J. Eriksson & Kirk P.H. Sullivan
IAFPA 2006 Factors that may affect speaker identification The listeners’… languge and dialect familiarity familiarity with the speaker age semantic expectation memory emotional state
IAFPA 2006 Language familiarity Hollien (2002): ’Most investigators have come to the conclusion that listeners find it difficult to identify talkers when they are speaking in a language other than their own’. Distracts the listener Subtle nuances in the language itself can be masked or lost Tend to reduce the number of available segmental constrasts What about Swedish listeners and their different dialectal background?
IAFPA 2006 A synthesis of results This presentation focus on the listeners’ semantic expectation, speaker familiarity, different dialect background and their age A study of one Swedish bidialectal male speaker Voice imitation of one Swedish politician –political speech –cooking speech - ’how to bake a cake’
IAFPA 2006 One Swedish bidialectal speaker The Princess and the Pea One male bidialectal Swede Four foils Listeners from Scania and Northern Sweden VoiceAgeDialect Target speaker 36Stockholm /Scanian Foil 127Scanian Foil 238Stockholm Foil 345Scanian Foil 425Stockholm
IAFPA 2006 The listeners Native speakers of Swedish from the region of Scania and the Northern Sweden 80 listeners – 50% male, 50% female –40 listeners from Scania 10 in each test –40 listeners from Northern Sweden 10 in each test
IAFPA 2006 Four voice line-up The composition of the voice identification tests TestFamiliarization voice Line-up voices SC-SCTargetSCFoil TargetSC ST-STTargetSTFoil TargetST ST-SCTargetSTFoil TargetSC SC-STTargetSCFoil TargetST SC = Scania, ST = Stockholm
IAFPA 2006 Results – all listeners Can the bidialectal speaker’s voice be identified?
IAFPA 2006 Results - dialect shifting tests Test 1: J2 training voice J1 target voice Inf2 – same dialect area as J2 Test 2: J1 training voice J2 target voice Inf3 – same dialect area as J1
IAFPA 2006 Results – listeners from North and South A regional difference? Control testsDialect shifting tests
IAFPA 2006 Imitation of one Swedish politican POLITICAL SPEECH – Professional imitation of Carl Bildt by AM –Carl Bildt (Swedish politician) –AM natural voice –4 other male adult speakers of Swedish COOKING SPEECH – Professional imitation of Carl Bildt by AM
IAFPA 2006 The listeners Exp 1 (political familiarisation passage) 128 adults 56 teens Exp 2 (cooking familiarisation passage) 121 adults 57 teens
IAFPA 2006 Results – the topic Political speechCooking speech
IAFPA 2006 Speaker familiarity Grp.Exp.No.Mean Age%FCB + NAdults NTeens CAdults CTeens SAdults STeens
IAFPA 2006 Results – speaker familiarity FCB+FCB- Political speech Cooking speech
IAFPA 2006 Results – different dialect background and age
IAFPA 2006 Speaker and dialect familiarity
IAFPA 2006 Discussion - sentence content Expectation –if you know the speaker/the target voice –if you use the same topic in the line-up as in the familiarization passage
IAFPA 2006 Discussion - familiarity with the speaker or the target voice Improve the results in a speaker recognition test A voice imitation often exaggerated for entertainment – and for recognition Support other studies in this area –among others; Hollien, Majewski & Doherty 1983, Yarmey 1995
IAFPA 2006 Discussion - familiarity with the dialect Familiarity with the dialect increase the identification rate… …but not in a dialect switch-condition (’native’ dialect) Listeners’ dialectal background seem to have an impact, especially if they are not familiar with the target voice
IAFPA 2006 Thanks! Acknowledgements: Funded by a grant from Band of Swedish Tercentenary Foundation Dnr K :1-4 to Umeå University for the project ’Imitated voices: A research project with applications for security and the law’.