Recognition of Speakers from Oral Non-Linguistic and Linguistic Sounds
Louise Owetz · Lund University Publications Student Papers (Lund University) · 2014
Voice recognition plays an important role in human communication and there is increasing interest in so called ‘earwitness’ testimony in the courtroom. However, research on voice identification or earwitness identification by way of short linguistic as compared to non-linguistic sounds remains in its infancy. To address this issue, the present study set out to examine speaker recognition from short linguistic and non-linguistic vocal sounds. A total of 45 participants were tested individually on a binary choice experiment to assess their ability to identify a previously unfamiliar voice when two other voices acted as lures. Participants first completed a familiarization session in which they heard either a male or female target voice describe their apartment and read a poem. This was followed by a session with 18 sounds from the target voice and two lures; each sound was repeated three times. The process was then repeated, with the target voice from the opposite sex. There was a difference in recognition abilities between one, two, and three-syllable words, where two and three-syllable words improved voice recognition. Performance with non-linguistic sounds was generally worse than that with linguistic sounds, but participants were able to identify target voices on the bases of sounds made by sighing, pondering, and hocking.