BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Generative models for audio and music processing - Taylan Cemgil\,
  Department of Engineering\, University of Cambridge
DTSTART:20070626T140000Z
DTEND:20070626T150000Z
UID:TALK7636@talks.cam.ac.uk
CONTACT:Oliver Williams
DESCRIPTION:The analysis of audio signals is central to the scientific und
 erstanding of human hearing abilities and in a broad spectrum of engineeri
 ng applications such as sound localisation\, hearing aids or music informa
 tion retrieval. Historically\, the main mathematical tools are from signal
  processing: digital filtering theory\, system identification and various 
 transform methods such as Fourier techniques. In recent years\, there is a
 n increasing interest for statistical approaches and tools from machine le
 arning. \nThe application of statistical techniques is quite natural: acou
 stical time series can be conveniently modelled using hierarchical signal 
 models by incorporating prior knowledge from various sources: from physics
  or studies of human cognition and perception. Once a realistic hierarchic
 al model is constructed\, many tasks such as coding\, analysis\, restorati
 on\, transcription\, separation\, identification or resynthesis can be for
 mulated consistently as Bayesian posterior inference problems. \nIn this t
 alk\, I will sketch our current work on audio and music signal analysis. I
 n particular\, I will illustrate various realistic generative signal model
 s such as factorial switching state space models\, Gamma-Markov random fie
 lds and point process models for music transcription\, restoration and sou
 rce separation. Some models admit exact inference\, otherwise efficient al
 gorithms based on variational or stochastic approximation methods can be d
 eveloped.\n
LOCATION:Small public lecture room\, Microsoft Research Ltd\, 7 J J Thomso
 n Avenue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR