BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:MPhil Project Practice Talks - Khuong Nguyen\, Duncan Roberts\, Yo
 rdan Zaykov\, James Snee
DTSTART:20110503T130000Z
DTEND:20110503T133000Z
UID:TALK30430@talks.cam.ac.uk
CONTACT:Andrew Rice
DESCRIPTION:Robot-based Evaluation of Bluetooth Fingerprinting\, Khuong Ng
 uyen\n\nIndoor tracking is an important topic for localisation. In this pr
 oject\, we implement a new\, efficient real-time system to locate an indo
 or object based on Bluetooth signal and fingerprinting technique. We espe
 cially tackle the hassle of collecting fingerprinting data by designing a
  robot to automate the whole process. Our robot can perform complex\, time
 -consuming data collection with great accuracy\, thus opening the door for
  various un-seen experiments. The project has three main goals:  We build
  a robot to collect data\;  We provide an in-depth analysis of the Blueto
 oth characteristic\, by answering research questions\;  We implement a com
 plete working Indoor Bluetooth tracking system for the Meeting room.\n\n\n
 Applying Language Models to Language Learning\, Duncan Roberts\n\n\nLearn!
  is a popular flash-card application for Android-based mobile phones cover
 ing everything from US Navy core values to English vocabulary. But flash-c
 ards are limited for language learning. With LearnGrammar!\, teachers crea
 te language learning tasks by providing a small number of sentences that e
 xemplify some grammatical point. The application generalises from the exam
 ple set\, providing students with many further examples and counter-exampl
 es to choose from in a multiple-choice question format.\n\nSpatial indexin
 g optimising of indoor localisation\, Yordan Zaykov\n\n Indoor tracking is
  an important topic for localisation. Recent work has made significant pro
 gress in this area\, but real-time localisation remains an open research p
 roblem. In this project we make a step forward in this direction by using 
 a precomputed spatial index of the building. We achieve a 14% speedup in t
 he localisation procedure and also reduce the precomputing stage to 1/10 o
 f the second using GPU programming. Current work on the project concentrat
 es on evaluating the trade-offs between reduced index size and decreased p
 erformance.\n\n\nTCP Performance and The Androids Sockets API\, James Snee
 \n\nRecent research into the power implications of the socket setup option
 s provided by the Android API has shown strange affects occurring under di
 fferent configurations. In an attempt to understand these affects a new ex
 perimental environment was built to measure and model the performance of t
 hese socket options. These models provide not only a deeper understanding 
 of the implications of socket setup in the Android Dalvik virtual machine 
 but can help developers build faster more efficient networked applications
 . \n
LOCATION:SS03\, William Gates Building
END:VEVENT
END:VCALENDAR