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:Inference Group
SUMMARY:A Cross-Entropy Based Method to Analyse Iterative
Decoding - Qinglin Luo
DTSTART;TZID=Europe/London:20051012T150000
DTEND;TZID=Europe/London:20051012T160000
UID:TALK4430AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/4430
DESCRIPTION:Iterative decoding provides a practical solution f
or the approaching of Shannon limit with acceptabl
e complexity. By decoding in an iterative fashion\
, the decoding complexity is spread over time doma
in while the overall optimality is still approacha
ble.\n \nEver since its successful application in
turbo codes in 1993\, people keep trying to discov
er the secrets behind iterative decoding. Till now
\, BER bounds\, density evolution\, EXIT chart\, G
aussian approximation are several most famous meth
ods that proves to be helpful for the analysis of
the behavior of iterative decoders. However\, rest
rictions like subject sequence must be Gaussian di
stributed\, transmitted sequence must be known\, a
pplicable region is either BER floor or BER clip o
nly\, etc.\, greatly limit the applications of the
se methods.\n\nIn this talk\, a new\, universal me
thod for the analysis of iterative decoding based
on cross-entropy will be discussed. We prove that
the maximum a posteriori probability (MAP) decodin
g algorithm minimizes the cross-entropy between th
e a priori and the extrinsic information subject t
o given coding constraints\, and the error correct
ing ability of each step of decoding can be evalua
ted with this cross-entropy for a converging turbo
decoder. Based on this proof\, the analysis of tu
rbo decoding on convergence rate\, derivation of E
b/N0 convergence threshold\, evaluation of error p
erformance in "error floor" region\, and design of
asymmetric turbo codes are carried out. Unlike mo
st conventional analysis methods which rely heavil
y on either Gaussian approximation of distribution
of the a priori/extrinsic information or a full k
nowledge of source bits\, or even both\, the new m
ethod provides analysis in a totally blind fashion
.
LOCATION:HEP Seminar Room\, Cavendish Laboratory
CONTACT:Phil Cowans
END:VEVENT
END:VCALENDAR