Burak Aydın May 2012 Wireless Sensor Networks - PDF

Burak Aydın May 2012 Wireless Sensor Networks Introduction Motivation-Background Mercury Architecture Case Studies Evaluation-Conclusion Wireless sensor networks are used in many areas such as military,environmental

Please download to get full document.

View again

of 41
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Self Improvement

Publish on:

Views: 51 | Pages: 41

Extension: PDF | Download: 0

Burak Aydın May 2012 Wireless Sensor Networks Introduction Motivation-Background Mercury Architecture Case Studies Evaluation-Conclusion Wireless sensor networks are used in many areas such as military,environmental monitoring,emergency,logistics etc. Additionally in healtchare and medical monitoring A wearable, wireless sensor platform for motion analysis of patients being treated for neuromotor disorders,such as Parkinson s disease and epilepsy. Designed to support long-term, longitudinal data collection on patients in hospital and home settings. Wireless sensor networks have the potential to greatly improve the study of diseases that affect motor ability In Mercury,a patient wears up to 8 sensors equipped with gyroscopes and accelerometers. A base station collects the data from nodes. Here are the key challenges: Energy consumption Battery lifetime Latency Accuracy Radio link conditions due to mobility Providing high-quality,clinically relevant data Approaches to challenges: Careful management of radio communications Management of flash storage and data processing Bandwith tuning by network Increased interest in body sensor networks for wearable applications as diverse as elder care Improved algorithms for processing data The basic approach is to capture triaxial accelerometer and gyroscope data from each limb segment (upper and lower arms and legs) using wearable sensors. 1 sensor 6 channels,100 hz,16-bit res bytes/sec data rate 2 Gbytes of flash capable of storing days of data Mainly, Parkinson s Disease and epilepsy diseases will be examined. The key diff. between these 2 is that epilepsy seizure detection should as fast as possible since this event is life-threatening. Mercury supports SHIMMER plaform( yet,not only SHIMMER) SHIMMER has: TI MSP430 Microcontroller MicroSD slot up to 2 GB of flash 250 mah Li-polymer battery is 10 gram,so easy to wear. The gyroscope consumes a large amount of energy,thus may be disabled when the node is not moving. Do not compute more computationally-demanding feautures,such as FFT, on the node itself Logging and rading features to flash energy saving use them Disabling gyroscope is good to increase lifetime However, negative impact on data fidelity which may or may not be acceptable on the application reqirements Mercury provides a balance between battery lifetime and data quality requirements A single app sampling,storage,feature comp.,reliable transfers Implemented on Pixie OS Activity filter Energy conservation Data Storage A sample block of 1200 bytes Feature extraction Sending huge amount of raw data is not energy efficient,hence compute features: maximum peak-to-peak amp. mean RMS peak velocity RMS of the jerk time series feature block is 600 bytes,plus it represents 30 sec worth of data in our case use it for energy saving Heartbeats Nodes periodically transmits to say that they are alive! Reliable transfer protocol-- end-to-end reliable transfer protocol,sensors sends data using ARQ The core of Mercury Runs on the base station and coordinates the WSN s operation Patients experience tremor,muscle rigidity,sluggish movements Intended to monitor fluctuations in patient s motor activity Standart driver makes no attempt to save energy,simply performs round-robin downloads Standart driver clinically-relevant data,yet cannot ensure the nodes meet a target lifetime Thus,different approaches than standart Throttling Downloads Download only if the node has excess energy i.e: e1(t) = C (C*t/t1),s=e1(t)- e(t) Download if s =0 Throttling gyro Disables the gyro sensor on nodes that are using so much energy Tradeoff is reduced data fidelity Other policies throttle sampling,throttle storage Rapid detection is important life-threatening if a seizure is suspected, the driver must first download the raw signal from all of the sensors worn by the patient battery depletion download only if preliminary indication of seizure A sensor alone can not detect aggregated information from multiple sensors Latency should be low hard to meet battery lifetime hence tell the nurse to replace the battery T: score threshold (whether an activity is seizure or not) k:is the required number of nodes to trigger a download cycle k=1 a single node may trigger a download cycle,causing many false results k=8 you may miss some subtle seizure activities To show Mercury has energy-saving features Parameters: 8 SHIMMER nodes Activity level of nodes between 25% and 100% Radio link quality:25% lossless, 50%lossy, 25%disconnected The driver drains the feature queue whenever there is good radio connectivity, while sample blocks are downloaded less frequently. The energy profile of the node varies based both on movement and download activity The throttle downloads driver saves additional energy, but does not do as well as the throttle gyro driver, which is effective at meeting the target lifetime. The throttle downloads and throttle storage drivers can obtain a peak lifetime of around 12 h. Throttle gyro does much better, with a peak of around 32 h, while throttling both gyro and downloads achieves nearly 42 h. Throttling both gyro and accelerometer scales almost linearly, but has a negative impact on data quality Aim was accurate detection,low latency 2 factors affect the performance noice and the choice of k Radio link varies across nodes being worn on different parts of the body As the noise increases beyond a certain amount, false triggers and limited bandwidth conspire to cause the network to download a large amount of non-seizure data, causing some seizure signals to be lost Using k = 8 performs better, as expected, since more spurious movements are filtered out yet some seizure events may be missed From a clinical perspective,seizures should be detected less than 5 minutes. The bar trace is the result when a real human activity is used rather than synthetic noise data The max. Lat is 170 sec, well below of 300 sec threshold In future,body area networks will be more frequently used than now. Mercury an important step to monior neurometer diseases in a home setting. Merc. Arch provides high data fidelity and achieve long lifetimes
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks