BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:(Research)  Energy Efficient Signal Acquisition in Wireless Sensor
  Networks: A Compressive Sensing Framework  / (Research) Particle filterin
 g on GPU for indoor pedestrian localisation - Wei Chen and Agata Bradjic
DTSTART:20110214T140000Z
DTEND:20110214T150000Z
UID:TALK28676@talks.cam.ac.uk
CONTACT:Andrew Rice
DESCRIPTION:Energy Efficient Signal Acquisition in Wireless Sensor Network
 s: A Compressive Sensing Framework\, Wei Chen\n\nWireless sensor networks 
 (WSNs) provide the ability to monitor various physical characteristics of 
 the real world\, such as sound\, temperature\, humidity\, etc.\, by distri
 buting a large number of inexpensive small devices in the detected environ
 ment. We present a novel approach based on the compressive sensing (CS) fr
 amework to monitor 1-D environmental information in WSNs. The proposed met
 hod exploits the compressibility of the signal to reduce the number of sam
 ples required to recover the sampled signal at the fusion center (FC) and 
 so reduce the energy consumption of the sensors in both sampling and trans
 mission. An innovative feature of our approach is a new random sampling sc
 heme that considers the causality of sampling\, hardware limitations and t
 he trade-off between the randomization scheme and computational complexity
 . In addition\, a sampling rate indicator (SRI) feedback scheme is propose
 d to enable the sensor to adjust its sampling rate to maintain an acceptab
 le reconstruction performance while minimizing the number of samples\, whi
 ch results in a reduced energy consumption in the sampling and transmissio
 n. A significant reduction in the number of samples required to achieve ac
 ceptable reconstruction error is demonstrated using real data gathered by 
 a WSN\nlocated in the Hessle Anchorage of the Humber Bridge.\n\nParticle f
 iltering on GPU for indoor pedestrian localisation\, Agata Bradjic\n\nA nu
 mber of approaches have been proposed over the years for localisation of p
 eople within the indoor environments. An approach that has been developed 
 in Computer Laboratory and that has evidenced a great promise uses inertia
 l sensing technology and a particle filter for eliminating the drift from 
 inertial sensors. The problem with this approach lies in the difficulty to
  scale particle filters to larger environments due to larger number of par
 ticles needed to perform the localisation. In this talk\, I will describe 
 how to attack this problem and improve scalability of the particle filter 
 approach by offloading particle filter processing onto a Graphics Processi
 ng Unit (GPU). I will discuss how I have dealt with different obstacles in
  GPU implementation and will present experimental results obtained for a l
 ocalization case study performed in the William Gates building.
LOCATION:FW11\, William Gates Building
END:VEVENT
END:VCALENDAR