WIRELESS SENSOR NETWORK (WSN) OUTLINE OF PRESENTATION Introduction to Wireless Sensor Network (WSN) and its Applications Introduction to Independent Component Analysis (ICA) Application of ICA in WSN
2 WIRELESS SENSOR NETWORK 3 A wireless sensor network (WSN) is a wireless network consisting of spatially distributed autonomous devices (called nodes/motes) using sensors to cooperatively monitor physical or environmental conditions, such as temperature, sound, vibration, pressure, motion or pollutants, at different locations. - Wikipedia APPLICATIONS Environmental monitoring Seismic activity detection; planetary exploration Industrial monitoring and control High-precision agriculture Structural health monitoring Social studies; healthcare and medical research Homeland security and military applications; surveillance, Detection of chemical/biological agents New areas keep emerging - Martin Haenggi, University of Notre Dame 4 APPLICATIONS - ENVIRONMENTAL MONITORING (1) 5 Cont 6 Transit Network Basestation Gateway Sensor Patch Patch Network Base-Remote Link Data Service Internet Client Data Browsing and Processing Sensor Node Cont
7 Parylene Sealant Acrylic Enclosures A project of Intel; UC, Berkeley and College of the Atlantic APPLICATIONS - ENVIRONMENTAL MONITORING (2) 8 APPLICATIONS - ENVIRONMENTAL MONITORING (3) 9 APPLICATIONS - AGRICULTURAL MONITORING 10 APPLICATIONS MEDICAL RESEARCH 11 APPLICATIONS SMART BUILDING 12 APPLICATIONS STRUCTURAL HEALTH MONITORING 13 APPLICATIONS SELF-HEALING MINEFIELDS 14 DESIGN ISSUES AND CHALLANGES Random deployment autonomous setup & maintenance Infrastructure-less networks distributed routing Energy, the major constraint trading off network lifetime for fault tolerance or accuracy of results Hardware energy efficiency Distributed synchronization Adapting to changes in connectivity Real-time communication, QoS Security
15 WSN SYSTEM OVERVIEW 16 WHAT IS A MOTE / NODE mote : noun something, especially a bit of dust, that is so small it is almost impossible to see - Cambridge Advanced Learners Dictionary Two main part Hardware Low Cost Low Power Tiny Software Lifetime maximization Robustness and fault tolerance Self-configuration
17 HARDWARE COMPONENTS OF MOTE Three main components A very low cost low power computer (typically a microcontroller) connected to one or more sensors A Radio Link to the outside world (plus Antenna) Power supply
18 MICROCONTROLLER DIVERSITY Flash based vs. SRAM based Combination of FLASH and CMOS logic is difficult Internal vs. External Memory Memory Size Digital Only vs. On-chip ADC Operating Voltage Range Operating Current, Power States and wake-up times Physical Size Support Circuitry Required External Clocks, Voltage References, RAM Peripheral Support SPI, USART, I2C, One-wire Cycle Counters Capture and Analog Compare Tool Chain Examples of microcontrollers being used in WSN are ATmega128, TI MSP430
19 SENSORS Photo-detector Temperature sensor 2D accelerometer Microphone (Acoustic threshold detector) Buzzer Magnetometer 20 RADIO LINK Commercially-available chips Available bands: 433 and 916MHz, 2.4GHz ISM bands Typical transmit power: 0dBm. Power control. Sensitivity: as low as -110dBm 21 POWER SOURCE AA batteries power the vast majority of existing platforms. They dominate the node size. Alkaline batteries indeed offer a high energy density at a cheap price. The discharge curve is far from flat, though. Lithium coin cells are more compact and boast a flat discharge curve. Solar cells are an option for some applications 22 Comparison of Energy Sources - UC Berkeley 23 Energy Management Issues Actuation energy is the highest Strategy: ultra-low-power sentinel nodes Wake-up or command movement of mobile nodes Communication energy is the next important issue Strategy: energy-aware data communication Adapt the instantaneous performance to meet the timing and error rate constraints, while minimizing energy/bit Processor and sensor energy usually less important 24 Energy Management Cont 25 EXAMPLES
WeC 99 Smart Rock MICA 26 EXAMPLES Cont MICA 27 - UC, Berkeley 28 SOFTWARE (OS) Examples include TinyOS, LiteOS, Contiki, ERIKA Enterprise, RIOT TinyOS is open source, event driven OS specifically designed for WSN When an external event occurs, such as an incoming data packet or a sensor reading, TinyOS signals the appropriate event handler to handle the event. Event handlers can post tasks that are scheduled by the TinyOS kernel some time later Current version is 2.1.2 releases in 2012 If you don't have mote hardware, you can compile the application for TOSSIM, the TinyOS simulator 29 GATEWAYS A gateway bridges a sensor network with another network, such as the Internet We can gather data on rainfall in a remote location and do the processing anywhere 30 Comparison with Ad Hoc Wireless Networks Both consist of wireless nodes but they are different. The number of nodes is very large Being more prone to failure, energy drain Not having unique global IDs Data-centric, query-based addressing vs. address- centric Resource limitations: memory, power, processing
32 Layer Architecture Application Layer Application models Target recognition, environment parameter detection and tracking Network Layer Routing layer models FLOODING, LEACH, SPIN, ADOV, ZRP Mac Layer Mac layer models TDMA, FDMA, 802.11 Physical Layer Physical layer models Sensing models, Battery models, Radio models 33 NS-2 Merits Event driven model Supports both wired and wireless Network animator Most widely used, a lot of documentation Demerits Comparatively difficult to learn and use, non-GUI Scalability problem No energy usage modeling 34 SIMULATION WORKFLOW IN NS-2 Topology definition: to ease the creation of basic facilities and define their interrelationships, ns-3 has a system of containers and helpers that facilitates this process. Model development: models are added to simulation (for example, UDP, IPv4, point-to-point devices and links, applications); most of the time this is done using helpers. Node and link configuration: models set their default values (for example, the size of packets sent by an application or MTU of a point-to-point link); most of the time this is done using the attribute system. Execution: simulation facilities generate events, data requested by the user is logged. Performance analysis: after the simulation is finished and data is available as a time-stamped event trace. This data can then be statistically analyzed with tools like R to draw conclusions. Graphical Visualization: raw or processed data collected in a simulation can be graphed using tools like Gnuplot, matplotlib or XGRAPH.
35 ICA AND BLIND SOURCE SEPARATION Blind signal separation, also known as blind source separation, is the separation of a set of source signals from a set of mixed signals, without the aid of information (or with very little information) about the source signals or the mixing process An example of Blind Source Separation is Cocktail Party Problem 36 COCKTAIL PARTY PROBLEM 37 Sources Observations s 1
s 2
x 1
x 2
Mixing matrix A x = As n sources, m=n observations Cont Independent component analysis (ICA) is a computational method for separating a multivariate signal into additive subcomponents by assuming that the subcomponents are non-Gaussian signals and that they are all statistically independent from each other. ICA is a special case of blind source separation - A.Hyvarinen, A.Karhunen, E.Oja, Independent Component Analysis
38 ICA ESTIMATION PRINCIPLES Principle 1: Nonlinear decorrelation. Find the matrix W so that for any i j , the components y i and y j are uncorrelated, and the transformed components g(y i ) and h(y j ) are uncorrelated, where g and h are some suitable nonlinear functions. Principle 2: Maximum nongaussianity. Find the local maxima of nongaussianity of a linear combination y=Wx under the constraint that the variance of x is constant. Each local maximum gives one independent component.
39 CENTRAL LIMIT THEOREM 40 APPLICATIONS OF ICA Applications include Blind source separation (Bell&Sejnowski, Te won Lee, Girolami, Hyvarinen, etc.) Image denoising (Hyvarinen) Medical signal processing: fMRI, ECG, EEG (Mackeig) Feature extraction, face recognition (Marni Bartlett) Time series analysis (Back, Valpola) Compression, redundancy reduction and most applications where Factor Analysis and PCA is currently used. While PCA seeks directions that represents data best in a |x 0 - x| 2 sense, ICA seeks such directions that are most independent from each other.
41 ICA VS PCA PCA Focus on uncorrelated and Gaussian components Second-order statistics Orthogonal transformation
ICA Focus on independent and non-Gaussian components Higher-order statistics Non-orthogonal transformation
42 FORMULATION OF ICA Assume that we observe n linear mixtures x1,...,xn of n independent components
we assume that each mixture xj as well as each independent component sk is a random variable The observed values xj(t), e.g., are the microphone signals in the cocktail party problem, are then a sample of this random variable In Vector-Matrix notation, above mixing model can be written as
xj = aj1*s1 + aj2*s2 + ... + ajn*sn, for all j X = As 43 FORMULATION OF ICA Cont In summation form, we can write
The 'mixing' matrix A is constant (a parameter matrix) The si are latent random variables called the independent components. Simply, we have to estimate both A and s, observing only x Estimation of A is performed under the following assumption about x The si are mutually independent The si are non-gaussian Under above assumptions, ICA estimation procedure can be as follows
44 FORMULATION OF ICA Cont 45 MEASURE OF GAUSSANITY 46 A SIMPLE EXAMPLE 47 A SIMPLE EXAMPLE Cont 48 A SIMPLE EXAMPLE Cont 49 EXAMPLE 2: IMAGE DENOISING 50 EXAMPLE3: SEPARATING TWO SINUSOIDS A = sin(linspace(0,50, 1000)); % A B = sin(linspace(0,37, 1000)+5); % B figure; subplot(2,1,1); plot(A); % plot A subplot(2,1,2); plot(B, 'r'); % plot B
figure; % compute and plot unminxing using fastICA c = fastica([M1;M2]); subplot(1,2,1); plot(c(1,:)); subplot(1,2,2); plot(c(2,:)); 51 APPLICATION OF ICA IN WSN 1. INTEGEGRATING WIRELESS EEGS INTO MEDICAL SENSOR NETWORK Electroencephalogram (EEG) is electrical recording of brain activity on different scalp positions. This recording is done in the form of potentials measured with electrodes. These potentials are assumed to be generated by underlying brain activities. Electrode recordings are assumed to be a mixture of temporal independent components and our task is to remove artifacts caused by muscular activities and blinking of eyes etc. ICA can be a promising technique in this regard.
52 Cont Modern trend is to connect human body monitoring devices using wireless sensor nodes called motes to form a wireless sensor network (WSN). Example of such WSN is to record EEG data digitized at 250 Hz per channel with 16 bit resolution resulting in 4 kbps. With a total of 24 such channels, total data rate can be 96 kbps which is in the supported range of Bluetooth (115.2 921.6). Benefit of using Bluetooth is that patient may be mobile. This data is then transmitted to PDA (which runs RTOS like TinyOS) using Bluetooth for later on upload to a server running shared machine intelligence algorithms for processing. The upload is done using TCP/IP connection with modified Media Access Control (MAC) protocol. Currently, only 7 channels have been employed. 53 54 55 56 Similarly eye blinking, horizontal and vertical movement, heartbeat etc. artifacts can be identified and removed Generally, bridges have been in service for 50-60 years and still they are expected to in service for 2020-30. Previously, bridge health was determined using hammering test, visual test or section test. Frequency of these tests is about 5 years. So, there is no mechanism for day to day health monitoring and diagnosis. It has been suggested to measure the vibration/displacement of bridge using sensor networks and associated signal processing units under the effect of external forces like impact load, wind, and vehicles crossing the bridge. Vibration effect of each load type may be separated from the observation data using Blind Source Separation technique like ICA. The idea is to simulate/measure the independent component of vibration due to each load type for a healthy bridge and then estimate the deterioration in bridge health using daily measurement data.
57 APPLICATION OF ICA IN WSN 2. BRIDGE DIAGNOSIS SYSTEM BY USING WSN AND ICA Cont When healthy bridge data is compared with a deteriorated one, its also observed that character frequency is found to be less in the bridge deteriorated by corrosion when vehicles were passed over them. Also character frequency is found to decrease as an effect of impact load and wind velocity over the aged bridge. 58 Cont 59 REFERENCES Aapo Hyvarinen: ICA (1999) http://www.cis.hut.fi/aapo/papers/NCS99web/node11.html ICA demo step-by-step http://www.cis.hut.fi/projects/ica/icademo/ Sarikaya, Behcet, M. Abdul Alim, and Siamak Rezaei. "Integrating wireless EEGs into medical sensor networks." Proceedings of the 2006 international conference on Wireless communications and mobile computing. ACM, 2006. Cheon, Jong-In, et al. "Development of Bridge Diagnosis Technology by Independent Component Analysis." SICE-ICASE, 2006. International Joint Conference. IEEE, 2006. Kim, Sukun, et al. "Health monitoring of civil infrastructures using wireless sensor networks." Information Processing in Sensor Networks, 2007. IPSN 2007. 6th International Symposium on. IEEE, 2007. The Neurophysiological Biomarker Toolbox (NBT) www.nbtwiki.net Feature extraction (Images, Video) http://hlab.phys.rug.nl/demos/ica/