|
||
    
|
Introduction Dr. Ashit Talukder is a
Associate Professor at the University of Southern California. He is also a Senior Researcher and Technical Project Manager at Childrens
Hospital Los Angeles/Univ. of Southern California and a Senior
Researcher (Category A) at the Jet Propulsion Laboratory/NASA,
California Institute of Technology, located in Pasadena, CA, USA. He
has a PhD from the
Dept. of Electrical and Computer Engineering, Carnegie
Mellon University where he worked
with Prof. David
Casasent. He obtained a M.S. at
Iowa State University where he worked with
Dr. Jennifer Davidson . He leads and works
on several projects funded by DARPA, NASA, commercial organizations,
and the NIH, in a variety of areas ranging from computer vision,
pattern recognition, and image processing, to intelligent sensor
networks, sensor fusion, data mining, and system optimization and
control.
 He has more than 40 publications in journals and conference proceedings. He is a member of the technical organizing committee of the Optical Pattern Recognition Conference in the Annual SPIE Defense and Security Symposium (formerly known as the Annual SPIE AeroSense Conference), held in Orlando, FL every year during April. He has chaired several conference sessions (including sessions on " Active vision in robotics " at Photonics East, Intelligent Robotics and Computer Vision XVII), and is a reviewer on several journals (IEEE Trans. Image Processing, IEEE Transactions on Signal Processing, IEEE Transactions on Systems, Man, and Cybernetics-B, Applied Optics, Neurocomputing, Neural Networks, and Optical Engineering).
MUSIC Project Description The sensor networks group at CHLA is led by Dr. Ashit Talukder ( talukder@usc.edu or Ashit.Talukder@jpl.nasa.gov). Our group builds complete hardware and software technologies for an end to end mobile monitoring system that will autonomously detect and reactively (adaptively) respond to events under extremely constrained bandwidth and power conditions. Towards this goal, we have built a wireless heterogeneous distributed sensor network with autonomous event detection and sensor processing capabilities, real-time multisensor control and resource (power, bandwidth, and storage) management. In MUSIC (Multi-modality Sensor network for Integrated event detection, C ontrol optimization and resource management), we address all issues encountered in mobile autonomous monitoring, including advanced algorithms for event and signal classification from multiple sensors, and adaptive power, bandwidth and storage management in mobile systems with limited resources. One of the applications that MUSIC is specifically being used for is mobile telehealth, where the health of individuals is autonomously monitored and reported in real-time using various physiological and metabolic measurements as the individual carries on his/her normal daily activities. One specific goal is to use MUSIC for mobile alcohol sensing monitoring and reporting for extended periods of time. The numeruous issues (and corresponding solutions) that we address in MUSIC for mobile telehealth is shown in the figure below.
Autonomous on-board sensing vs. Data Transmission: Why current Sensor Networks are not enough. Current sensor networks are designed for observing and capturing environmental data via sensory observations and transmitting such data back to a central location for manual analysis. However, data transmission is a power-hungry operation. In our tests, we have found that data transmission consumes about 6-10 times more power than data processing. In addition, autonomous processing paves the way for autonomous sensing systems where events can be detected "in-situ " without having a person to look at the data. This reduces bandwidth usage, increases sensor network lifetime, and reduces the need for manual intervention. It also raises the possibility of autonomous "bold"-reactive systems, where the system can autonomously take corrective action when it detects an event. This could be useful in:
MUSIC IN ACTION We have designed a generic wireless sensor network system using Motes RF nodes with interfaces to COTS and specialized sensors, in a three-layered structure that allows for breakdown at any node without creating a catastrophic failure in other layers and nodes. Each node is equipped with autonomous processing capability and adaptive resource management algorithms that allow autonomous operation during communication link failures. The figure below shows an individual wearing the MUSIC sensor network system and a screenshot of a wireless pulse oximeter sensor node.
| ||||||||||||||||||||
This site was last updated 12/09/2004
