BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY: Modeling Natural Sounds with Modulation Cascade Processes - Richa
 rd Turner\, Gatsby Neuroscience Unit\, UCL
DTSTART:20070510T120000Z
DTEND:20070510T130000Z
UID:TALK7040@talks.cam.ac.uk
CONTACT:Taylan Cemgil
DESCRIPTION:Auditory scene analysis is extremely challenging. One approach
 \, perhaps that adopted by the brain\, is to shape useful representations 
 of sounds on prior knowledge about their statistical structure.\nFor examp
 le\, sounds with harmonic sections are common and so time-frequency repres
 entations are efficient. Most current representations concentrate on the s
 horter components. Here\, we propose representations for structures on lon
 ger time-scales\, like the phonemes and sentences of speech. We decompose 
 a sound into a product of processes\, each with its own characteristic tim
 e-scale. This demodulation cascade relates to\nclassical amplitude demodul
 ation\, but traditional algorithms fail to realise the representation full
 y. A new approach\, probabilistic amplitude demodulation\, is shown to out
 -perform the established methods\, and to\neasily extend to representation
  of a full demodulation cascade. Finally we outline how to wed this model 
 to one that models the dynamic Fourier components of sounds in order to pr
 oduce a full generative model for natural sounds.\n
LOCATION:LR5\, Engineering\, Department of
END:VEVENT
END:VCALENDAR