Project Background
The TRUST SUPERB-IT projects are part of the TRUST EMR Project and the ITALH project. An overview of these projects can be found in this presentation [pdf]
Implementation of an Electronic Medical Record System
Sonny Hernandez, Mentor: Arsalan Tavakoli
With movements from paper to electronic formats for patient medical records which is happening now, and the maturation of wireless sensor systems come the possbility of integrating continuous monitoring of patient biometric and other information into medical records. This could provide doctors with previously impossible to obtain but very useful outpatient and in-home data to more fully diagnose their patients and monitor their progress. However, many technical challenges remain to be solved in order to achieve this.
In this project, the student will implement an open source EMR package locally and extend it to include automatic input from sensor systems. The form of the data stored will be investigated in terms of statistics, features, etc. that can be used to capture the important information without storing the full data set.
Design of a Distributed Tracking System for Camera Networks
Jessica Jimenez Pellot, Mentor: Edgar Lobaton
Tracking is one of the main applications of sensor networks. Nowadays there are networks of cameras dedicated to this purpose, which combine all of their visual information in centralized schemes. However, there is a need for having decentralized systems capable of pre-processing data for each camera node in order to reduce the network traffic, and increase privacy (especially important in wireless networks). Of course, it is also of interest to integrate this network of visual information with other sensory information.
The objective of this project is the design of a tracking algorithm for a sparse camera network using a distributed scheme while trying to minimize the flow of information between the different camera processes, and its integration with other modes of sensory input. As part of the project we will work on interfacing the cameras with a sensory network composed of motes. The first stage of the implementation will involve the simulation of the distributed process in a centralized server, and then aim for implementation on a network of distributed hardware resources.
Camera Networks and Computer Vision
Jamie Lauren Webb, Mentor: Marci Meingast
This project focuses on designing and running a small-scale camera network. The
initial phase of the project will involve setting up a camera network consisting of 6-12
Logitech webcams. There will be a couple important aspects to this phase. Looking at
the field of view of each camera and designing the network so that the cameras will cover
the desired area is the first step. Once this is done, it will also be important to see what
the accuracy of 3D construction will be with this setup. Syncing up the cameras so that
they take images at the same time and building a simple interface to turn the network on
and off will be the final step.
The next phase will involve doing tracking tasks. A colored marker placed on a
wearable device on a human subject will be tracked. Designing the algorithm and
looking at how to automate which camera images are necessary for tracking at a given
time step will be important. Once a colored marker can be tracked, the tracking task will
be extended to tracking a human with no without any pre-defined marker in place.
Tracking will be based on background subtraction and motion detection. The tracking
can then be further extended to deal with multiple humans and deal with occlusion and
cross-over issues.
The goal is that this sort of camera network can be placed in the home to do some
low level monitoring of people with health concerns. The information from the camera
network could be used to send alerts to a healthcare facility when an emergency arises.
Empirical Robustness Analysis of Wireless Connectivity in Sensor
Network Deployments
Tonmoy Bhattacharjee, Mentor: Phoebus Chen
In this project, the student will conduct experiments and develop
software tools for characterizing the robustness of the communication
between wireless sensor nodes in a sensor network deployment. The
experiments will be conducted in different settings -- indoors,
outdoors, in cluttered environments -- with an emphasis on developing
guidelines for the deployment of sensor networks for health care
systems. In sensor network deployments for health care, the emphasis
is on creating a redundant, robust communication backbone from static
sensor node placements that can provide good coverage to mobile sensor
nodes worn on the health care patients. The focus on building robust
wireless networks is to minimize the cost and inconvenience of
installing wiring to connect all the sensor nodes.
The project has three levels of deliverables. At the most basic
level, the student will learn to program and use the tools in our
sensor network programming environment, TinyOS, to perform empirical
studies of the wireless connectivity of sensor network deployments in
typical household settings and outdoor environments. The student will
provide a set of basic guidelines on the spacing and placement of
nodes to achieve acceptable levels of communication robustness. The
next level of deliverables is for the student to develop a wireless
handheld diagnostic device, such as a PDA connected to a wireless
sensor node, that can be used for rapidly characterizing the wireless
connectivity in a static multipath environment and help with sensor
node deployments. For the ambitious student, the last level of
deliverables would be to study the performance of different
networking/routing protocols for health care sensor network
deployments, and understand how the parameters in these protocols
should be tuned for different deployments.
Statistical Behavior Models of Human Activities
Kaseima Frye, Mentor: Songhwai Oh
In this project, the student will build a statistical behavior model of
an elderly person living alone. The intern will
research what types of sensor can be used to monitor the behavior of
an elderly person living alone;
collect sensor data from an existing sensor network deployment;
develop sensor models;
build different statistical behavior models of an elderly person
living alone; and
evaluate different statistical behavior models.
In order to build a statistical model, we will use some techniques from
machine learning such as Bayesian networks and unsupervised learning. In
particular, we will first apply the hidden Markov model (HMM) to model
spatio-temporal behavior patterns and use the expectation-maximization
(EM) method to learn the parameters of the model. Strong background in
discrete probability is required and familiarity with MATLAB is preferred.
Time Synchronization Security in Sensor Networks
Jocelyn Adams, Mentor: Tanya Roosta
Time synchronization is one of the fundamental tasks in sensor networks, and many applications, such as transmission scheduling among other things, heavily rely on accurate time synchronization. The purpose of time synchronization is to give a unified view of the time to all the nodes in the network. Different nodes have different clocks, and the frequency of these clocks changes over time due to the quartz crystal quality, temperature, as well as other factors. Therefore, there is a need to periodically update each node’s local time and make sure it is consistent with other nodes’ view of the time. There has been a number of time synchronization protocols proposed for sensor networks, however, none of them take security into account when designing the protocol. It is easy for an adversary to take over a few nodes and inject faulty time updates into the network. As a result, the sensor network will become out of sync, and will not be able to function properly.
The purpose of this project is to implement one of the main time synchronization protocols that has been proposed, i.e. Flooding Time Synchronization Protocol, on a real test-bed, in order to quantitatively evaluate the effect of faulty time updates on time synchronization protocol. Furthermore, we would like to show the effect of the attacks on time synchronization on surveillance application which is one of the most important applications of the sensor network.
Downloads
|
