Você está na página 1de 3

FYP Proposal

Student: Patrick Brady No: 03861783 Class: 4th Year Electronic & Computer Engineering Project Title: Fall detection system for elderly patients

Overview
Falls have come to be recognized as a major threat to the safety, health and independence of elderly persons. For final year project I propose to research and develop a complete solution for detecting falls in elderly people. The system will detect when a patient has fallen and issue an alert to a third party, e.g. to notify a nurses station in a home for the elderly, notify relatives, notify a GP. Because a fall is usually a random event it is difficult to determine when a fall has occurred. But if you could determine, based on current sensor readings, the likelihood (to a high degree of probability) that a fall has occurred, then this could be used to infer the probability of this event and issue an alarm. The core problem that my system will solve is to what degree of accuracy falls can be detected. With this in mind I propose to model the random event (i.e. the fall) using a Bayesian network. A Bayesian network is a mathematical formalism based on probability theory that allows us to reason about beliefs under conditions of uncertainty [1]. To sense the event of a fall, the model will incorporate several T-Mote Invent sensors attached to the patients body [2]. T-Mote Invent sensors have accelerometers built in to them that can be used to detect fast changes in motion. As part of this project, I will need to program the sensors using an embedded operating system such as TinyOS [3]. The sensor readings would be used as the current observations for the Bayesian network. The readings from the sensors will be transmitted wirelessly to a base station connected to the serial port on a host computer running GSN [4]. GSN (Global Sensor Networks) is a Java-based middleware engine that allows to store and process sensor data. To capture and process the live data from the sensors, I will have

to write a wrapper (a kind of device driver) and several java classes to implement the Bayesian network. Once I have implemented the Bayesian network, I will test it with the sensors and measure the accuracy of the system in detecting falls in realistic conditions. Once the system response and accuracy is adequate, I will write a Java based GUI to implement the alerts, for example, Email, SMS or actual phone call notifications. If time permits, I will integrate my system with an Electronic Healthcare Record system to allow GPs to monitor the health of their elderly patients and be notified of falls events.

Work Plan and Schedule


Phase1: The initial phase of this project will consist of programming the T-Mote sensors using nesC and TinyOS. Phase2: The second phase will be to develop the Bayesian algorithm and to integrate the sensor data with this algorithm. -Phases 1 and 2 to be completed by the end of December 2008/early January 2009 Phase3: Integrate and test the sensor data with GSN -Phase3 to be completed by the end of February 2009 Phase4: Implement notification component and complete project report. -Phase 4 to be completed by the end of March 2009 Project report will be outlined throughout entire project.

Materials and Tools


1. Hardware: Tmote sensors. 2. Software: GSN, Java SDK, TinyOS, nesC. 3. Other: Library resources on Bayesian networks, book on sensor networks.

References
[1] [2] Bayesian Network, http://en.wikipedia.org/wiki/Bayesian_network T-Mote Invent data sheet, http://www.sentilla.com/pdf/eol/tmote-invent-user-guide.pdf

[3] [4]

TinyOS Operating System, http://www.tinyos.net/ Global Sensor Networks, http://gsn.sourceforge.net/

Você também pode gostar