BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Probabilistic Amplitude and Frequency Demodulation - Dr Richard Tu
 rner\, Machine Learning Group\, CUED
DTSTART:20120425T131500Z
DTEND:20120425T140000Z
UID:TALK37260@talks.cam.ac.uk
CONTACT:Rachel Fogg
DESCRIPTION:A number of recent scientific and engineering problems require
  signals to be decomposed into a product of a slowly varying positive enve
 lope and a quickly varying carrier whose instantaneous frequency also vari
 es slowly over time. Examples include the analysis of speech signals and b
 rain imaging signals\, like EEG. Although signal processing provides\nalgo
 rithms for so-called amplitude- and frequency-demodulation\, there are wel
 l known problems with all of the existing methods. \n\nMotivated by the fa
 ct that amplitude and frequency demodulation is ill-posed\, we approach th
 e problem using probabilistic inference. The new approach\, called probabi
 listic amplitude and frequency demodulation (PAFD)\, models instantaneous 
 frequency using an auto-regressive generalization of the\nvon Mises distri
 bution\, and the envelopes using Gaussian auto-regressive dynamics with a 
 positivity constraint. A novel form of expectation propagation is used for
  inference. We demonstrate that\nalthough PAFD is computationally demandin
 g\, it outperforms previous approaches on synthetic and real signals in cl
 ean\, noisy and missing data settings.\n
LOCATION:LR10\, Engineering\, Department of
END:VEVENT
END:VCALENDAR