BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Signal Processing and Communications Lab Seminars
SUMMARY:Demodulation and time-frequency analysis as infere
nce - Dr Rich Turner\, CUED.
DTSTART;TZID=Europe/London:20130530T140000
DTEND;TZID=Europe/London:20130530T150000
UID:TALK45602AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/45602
DESCRIPTION:In this talk I will present a theoretical framewor
k that links a set of widely used methods from sig
nal processing to statistical inference procedures
. This result will then be used as a conceptual sp
ringboard \nto improve upon the classical methods.
\n\nI will begin by describing a family of related
inference problems that have optimal solutions co
rresponding to the short-time Fourier transform (S
TFT)\, spectrogram\, filter bank\, and wavelet \nr
epresentations of signals. The framework allows us
to use modern techniques from statistical inferen
ce to improve upon these classical signal processi
ng methods.\n\nI will show two examples where such
an approach has borne fruit. In the first example
we use an inferential \nextension of the Hilbert
method to produce high-quality approaches to joint
amplitude and frequency modulation of signals. Th
e new approach is uncertainty-aware and therefore
noise robust which results in fewer \nartifacts. I
n the second example we extend the STFT into an ad
aptive time-frequency analysis using a hierarchica
l probabilistic model. The parameters of the new r
epresentation\, including the channel \ncentre-fre
quencies and bandwidths\, can be learned directly
from the signal. The adaptive representation can b
e used to remove noise from signals and to impute
missing data. Surprisingly\, the method is an \nex
cellent model for naturally occurring audio textur
es such as howling wind\, falling rain\, and runni
ng water.\n\nI will wrap up by discussing how we m
ight bring the fields of signal processing and sta
tistical inference closer together and the benefit
s and challenges of such a research effort.\n
LOCATION:LR11\, Engineering\, Department of
CONTACT:Dr Ramji Venkataramanan
END:VEVENT
END:VCALENDAR