lecture13

lecture13 - Disguise, drunkenness, lying, and other random...

Info iconThis preview shows pages 1–14. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Disguise, drunkenness, lying, and other random topics vaguely related to speaker recognition Where we left off Recognizing familiar voices is different (at the level of brain organization) from discriminating among unfamiliar voices. Recognizing familiar voices seems to require mostly pattern recognition skills. Remembering and/or discriminating among unfamiliar voices requires both pattern recognition and featural analysis. How do we study memory for unfamiliar voices? Early approaches: lists of features So many features! Factor analysis: reduces a large list of features to a small set of independent factors Limitation: Garbage in, garbage out Deriving a valid set of features Multidimensional scaling Results depend on input set of voices Many discrepancies from study to study General limitation of feature-based approaches: Evidence reviewed last time suggests that this isnt the way we remember or recognize voices. Prototypes and features How many prototypes? What kinds? Who knows? A few acoustic features turn up consistently: F0 (mean, intonation pattern, creakiness, breathiness, etc.) Formant frequencies Loudness Why F0? Earliest information we have about voice Ears are sensitive to changes in frequency. F0 varies with physiology so its a good signifier. F0 reflects changes in respiratory status, neural activity/arousal, and articulation. We can control F0 very finely. Why F0? Frequency information is redundantly represented in the acoustic signal, so its robust against noise, hearing loss, etc. F0 doesnt interact too much with the spoken message (at least not in English). Additional limitations to these studies Almost all work done on speakers of English More than of the worlds languages use F0 to carry word-level meaning (Thai, Chinese) No information is available about how this affects the importance of F0 as a carrier of speaker-specific information. Studies of tone in Thai aphasics What kinds of factors affect speaker recognizability in forensic situations? Speaker characteristics Characteristics of the voice sample Transmission system characteristics Listener factors Factors related to the way recognition is measured 1. Effects of speaker characteristics Voices differ in how recognizable they are Familiarity Language spoken/language typology (speakers of English can recognize speakers of German better than can speakers of Spanish or Chinese, but worse than can speakers of German). More speaker characteristics: Disguise and mimicry An impersonator identifies the vocal features that characterize a speaker and copies those features....
View Full Document

Page1 / 81

lecture13 - Disguise, drunkenness, lying, and other random...

This preview shows document pages 1 - 14. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online