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:Gaussian Particle Implementations of Probability H
ypothesis Density Filters - Daniel Clark
DTSTART;TZID=Europe/London:20070222T130000
DTEND;TZID=Europe/London:20070222T140000
UID:TALK6693AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/6693
DESCRIPTION:The Probability Hypothesis Density (PHD) filter is
a multiple-target filter for recursively estimati
ng the number of targets and their state vectors f
rom sets of observations.\nThe filter is able to o
perate in environments with false alarms and misse
d detections. Two distinct algorithmic implementat
ions of this technique have been developed. The\nf
irst of which\, known as the Particle PHD filter\,
requires clustering techniques to provide target
state estimates which can lead to inaccurate estim
ates and is computationally expensive.\nThe second
algorithm\, called the Gaussian Mixture PHD (GM-P
HD) filter does not require clustering algorithms
but is restricted to linear-Gaussian target dynami
cs\, since it\nuses the Kalman filter to estimate
the means and covariances of the Gaussians. Extens
ions for the GM-PHD filter allow for mildly non-li
near dynamics using extended and Unscented Kalman
filters. \nA new particle implementation of the PH
D filter which does not require clustering to dete
rmine target states is presented. \nThe PHD is app
roximated by a mixture of Gaussians\, as in the GM
-PHD filter but the transition density and likelih
ood function can be non-linear. \nThe resulting fi
lter no longer has a closed form solution so Monte
Carlo integration is applied for approximating th
e prediction and update distributions. \nThis is c
alculated using a bank of Gaussian particle filter
s\, similar to the procedure used with the Gaussia
n sum particle filter. \nThe new algorithm is deri
ved and presented with simulated results.\n
LOCATION:LR6\, Engineering\, Department of
CONTACT:Taylan Cemgil
END:VEVENT
END:VCALENDAR