Wavelet-Based Data Compression Technique for Wireless Sensor Networks

Transcription

1 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Wavelet-Based Data Compressio Techique for Wireless Sesor Networks P. Kumsawat, N. Pimpru, K. Attakitmogcol ad A.Srikaew Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 Abstract I this paper, we proposed a efficiet data compressio strategy exploitig the multi-resolutio characteristic of the wavelet trasform. We have developed a sesor ode called Smart Sesor Node; SSN. The mai goals of the SSN desig are lightweight, miimal power cosumptio, modular desig ad robust circuitry. The SSN is made up of four basic compoets which are a sesig uit, a processig uit, a trasceiver uit ad a power uit. FiOStd evaluatio board is chose as the mai cotroller of the SSN for its low costs ad high performace. The software codig of the implemetatio was doe usig Simulik model ad MATLAB programmig laguage. The experimetal results show that the proposed data compressio techique yields recover sigal with good quality. This techique ca be applied to compress the collected data to reduce the data commuicatio as well as the eergy cosumptio of the sesor ad so the lifetime of sesor ode ca be exteded. Keywords Wireless sesor etwork, wavelet trasform, data compressio, ZigBee, skipped high-pass sub-bad. W I. INTRODUCTION IRELESS sesor etworks (WSN) is a self-orgaized distributed itelliget system comprisig low-cost, lowpower, multifuctioal sesor odes that are small i size ad commuicate with each other i short distaces. The developmet of such etworks was origially motivated by military applicatios such as battlefield surveillace. Recetly, a lot of research related WSN have bee coducted ad people have bee realizig theirs ulimited applicability. For example, WSN ca be used for data collectio purposes i situatios such as eviromet ad habitat moitorig, healthcare applicatios, home automatio, structural moitorig, ad equipmet diagostics. However, WSN face may challeges, maily caused by commuicatio failures, limited storage capability ad computatioal costraits ad limited power supply. Therefore, the techology of WSN is requirig more extesive research ad developmet before it becomes practical. I previous work, Watthaawisuth et al. [1] proposed a GPS tractor trackig system usig ZigBee multi-hop mesh etwork for data commuicate i the farm. The system ca help farmer for maagig ad reducig the resources for tractor or other vehicles i the farm. P. Kumsawat ad N. Pimpru are with the School of Telecommuicatio Egieerig, Istitute of Egieerig, Suraaree Uiversity of Techology, Nakho Ratchasima, 000, Thailad (phoe: ; fax: ; prayoth@ sut.ac.th). K. Attakitmogcol ad A. Srikaew are with the School of Electrical Egieerig, Istitute of Egieerig, Suraaree Uiversity of Techology, Nakho Ratchasima, 000, Thailad. Data compressio is a process that reduces the amout of data i order to reduce data trasmitted ad/or decreases trasfer time. Due to the limited processig ad storage resources of the sesor odes, data compressio i sesor odes requires the simple ad lightweight algorithms. I recet years, people have doe a lot of research work o data compressio algorithm ad proposed may compressio algorithms for WSN []. I [3], Kulakov ad Davcev proposed data acquisitio through hierarchical two-level architecture with algorithms usig wavelets for iitial data-processig of the sesory iputs ad eural-etworks usig usupervised learig for categorizatio of the sesory iputs. This architecture provides a big dimesioality reductio ad i the same time additioal commuicatio savig, sice oly classificatio IDs (small biaries) are passed to the cluster head istead of all iput samples. I [4], Gohet al. proposed the compressio of euroal recordigs i real-time usig a ovel discrimiatig Lide-Buzo-Gray algorithm (DLBG) that preserves spike shapes while filterig backgroud oise. The techique is implemeted i a low power digital sigal processor (DSP) which is capable of wirelessly trasmittig raw euroal recordigs. Depedig o the sigal to oise ratio of the recordig, the compressio ratio ca be tailored to the data to maximally preserve power ad badwidth. I [5], Nasri et al. proposed a sigal compressio approach i WSN cosistig of techique to skip computatio of certai highpass coefficiets of the discrete wavelet trasform (DWT) called SHPS (Skipped High-Pass Sub-bad). The simulatio results show that the proposed scheme optimizes etwork lifetime, reduces sigificatly the amout of required memory ad computatio eergy. I [6], Kimura ad Latifi proposed five differet types of data compressio schemes which have bee specifically desiged for WSN: codig by orderig, pipelied i-etwork compressio, JPEG000, lowcomplexity video compressio, ad distributed compressio. Eve though those compressio schemes are still uder developmet, experimetal results idicate that their compressio rate ad power reductio maers are quite impressive. I [7], Chichi et al. proposed ew data compressio algorithm ispired from Ru Legth Ecodig called K-RLE. The authors also evaluate ad compare compressio algorithms o a ultra-low power microcotroller from Texas Istrumet withi the MSP4 series used for desigig wireless sesor etworks. Recetly, the wavelet-based approach has attracted much attetio from researchers due to simplicity ad high compressio performace. I [8], Nasri et al. proposed a alterative image trasmissio approach i WSNs, based o Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

2 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 JPEG000 image compressio stadard. This approach is based o discrete wavelet trasform (DWT) ad embedded block codig with optimized trucatio (EBCOT) which uses a better order of trasmissio. Performace of the proposed image compressio scheme is ivestigated with respect to image quality ad eergy cosumptio. I [9], Mahas et al. proposed a efficiet image trasmissio strategy exploitig the multi-resolutio characteristic of DWT. The authors use selective decodes-adforward (SDF) cooperatio. The experimetal results show that the overall eergy cosumptio ca be cosiderably reduced, with a egligible decrease o the average image quality. I [10], Rajput et al. provided a brief survey of advatage of usig WSN i agriculture. The challege i usig WSN for apple farmig is discussed. I [11], Kohvakka et al. proposed a mathematical performace aalysis ad simulatios of IEEE LR-WPAN i a large-scale WSN applicatio with up to 1560 odes. The etwork is formed i a beaco eabled cluster-tree topology accordig to ZigBee specificatio. The performace of a device ad a coordiator are aalyzed i terms of the average power cosumptio ad throughput. I [1], a extesive survey of computatioal itelligece applicatios to various problems i WSN from various research areas ad publicatio veues is preseted i the paper. I this paper, we propose a sigal compressio approach i WSN, based o wavelet trasform called SHPS. This approach does ot require computig high-pass coefficiets i order to reduce the umber of executed operatios ad therefore save computatio eergy used durig the wavelet compressio process. We have developed embedded software implemetatios based o the compressio algorithm described i [5]. Due to its simplicity, this algorithm is very fast ad ca be easily implemeted. The computatioal blocks i these realizatios are implemeted i the Simulik model ad MATLAB programmig laguage. We also implemet the hardware of the smart sesor ode usig various compoets. II. PRELIMINARIES A. Wavelet Trasform The wavelet trasform has received a tremedous amout of iterest i may sigal processig ad image processig applicatios. The priciple objective of the wavelet trasform is to hierarchically decompose a iput sigal ito a series of successively lower frequecy approximatio sigal ad their associate detail sigal. Suppose φ (t) ad ψ (t) are the scalig fuctio ad the correspodig wavelet respectively with fiite support [ 0, l ], where l is a positive umber. It is well kow that φ (t) ad satisfies the followig dilatio equatio: ψ ( t) = g( k) ϕ( t k) () k= where the h(k) ad g (k) are costats called low- ad highpass filter coefficiets, respectively. We will use the followig stadard otatios: j / j φ j, k ( t) = φ( t k) (3) j / j ψ j, k ( t) = ψ ( t k) (4) Each wavelet ψ j, k ( t) is geerated by traslatig ad dilatig of fuctio ψ (t) called mother wavelet. Cosider the subspace ad the subspace V j of L defied by; V = Spa { ϕ, k Ζ } (5) j W j of j, k L defied by W = Spa { ψ, k Ζ } (6) j ' the subspaces V j, < j <, form a multi-resolutio of with the subspace V j+1. I fact, the as; j, k W j beig the differece betwee L V j ad L space has or the ormal decompositio L = V j W j (7) j I most practical applicatios, oe ever explicitly calculates the scalig fuctio φ (t) ad wavelet fuctio ψ (t) but performs the trasform usig the scalig coefficiets h(k) ad the wavelet coefficiets g (k). I forward wavelet aalysis, a J -level discrete decompositio ca be writte as f ( t) = c ϕ( t ) 0, = c ϕ ( t) + d ψ ( t) j, k j, k j, k j, k k j= 1 k J where the coefficiets c j, k ad d j, k at resolutio j are related to the coefficiets c j 1, k at level j 1by the followig recursive equatios: (8) ad ϕ( t) = h( k) ϕ( t k) k= (1) c j, k = c j 1, kh( k) (9) d j, k = c j 1, k g( k) (10) Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

3 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 for j = 1,,..., J. I (8), the first summatio gives a fuctio that is a low resolutio or coarse approximatio of f (t), which represets the smooth part of f (t). For each icreasig level j i the secod summatio, a higher or fie resolutio fuctio is added, which represets the detail part of f (t). I backward wavelet sythesis, a recostructio of the origial fie scale coefficiets of the sigal ca be made from a combiatio of the scalig coefficiets ad wavelet coefficiets at a coarse resolutio. Because all of these fuctios are or the ormal, we have c j 1, k = c j, h( k ) + d j, k g( k ) (11) The sythesis operatio of (11) is equivalet to up-samplig the coefficiets c j+ 1, ad d j+ 1, d i the coarser level j + 1, ad the covolvig with h (k) ad g (k), idividually, to obtai the scalig coefficiets i the fier level j. The sythesis process ca be recursively cotiued to the origial level. The aalysis ad sythesis procedures lead to the pyramid-structured wavelet decompositio [13]. The block diagrams of a wavelet decompositio ad recostructio are show i Fig. 1, where H(z) ad G (z) are the z trasform of h (k) ad g (k), respectively. (a) (b) Fig. 1The block diagrams of two-level wavelet (a) decompositio ad (b) recostructio B. ZigBee ZigBee is a specificatio for a suite of commuicatio protocols based o the IEEE stadard, which targets wireless persoal area etworks. We ote that ZigBee operates i the etwork layer usig as a trasport layer. These services are provided by the IEEE protocol. Actually, the IEEE defies two physical layers which operate i three frequecies. The physical layer operates at.4 GHz with a maximum trasfer rate of 50 kbps. There are o restrictios for usig this bad aroud the world. There exist three types of ZigBee devices as follows: The ZigBee Coordiator (ZC): This is the most powerful ZigBee device. The coordiator ca be see as the root of the etwork topology ad it ca also be utilized as a gateway to other picoets. Sometimes, the coordiator is used as a trust etity that ca maitai the system s key repository. ZigBee Router (ZR): This device ca execute a commo applicatio ad ca work as itermediate router i order to sed data to other ZigBee devices. ZigBee Ed Device (ZED): This device has limited fuctioalities such as exchagig iformatio with the ZC or the ZR devices. A ZED caot forward data to other devices. The mai feature of this device is that it keeps the device stay i the low-power cosumptio mode most of the time. This allows the savig of sigificat battery life time. This device requires less amout of memory ad is the cheapest ZigBee device. Example of a etwork of ZigBee devices ad the compariso of several wireless sesor etwork techologies are show i Figs. ad 3, respectively. Fig. Example of a etwork of ZigBee devices Fig. 3 Compariso of several wireless sesor etwork techologies III. PROPOSED METHOD I this sectio, we first give a brief overview of the WSN embedded system architecture ad sigal compressio algorithm i the wavelet trasform domai. The mai motivatio of our work is based o the idea proposed i [5]. The computatioal blocks i these realizatios are implemeted i the Simulik model ad MATLAB Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

4 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 programmig laguage. We also implemet the hardware of the WSN usig various compoets. A. WSN Embedded System Architecture Fig. 4 shows diagram of a wireless mesh etwork. Mesh is oe of may etwork topologies ad forms iheretly the most reliable ad scalable etwork. Each ode has routig capabilities ad ot oly passes o packets but also decides which is the best path ad igores ay broke odes. I Fig. 5, the mai cotroller of the SSN is the FiO Std evaluatio board [14]. The microcotroller uit (MCU) is ARM 3-bits Cortex TM M3 processor (STM3F103RET6). It cotais also read oly memory (ROM), radom access memory (RAM), a 1 bit aalog to digital covertor (ADC), a 1 bit digital to aalog covertor (DAC), timer ad few comparators. We have itegrated sesors such as relative temperature ad humidity sesor, light sesor, soil moisture sesor to the FiO Std evaluatio board. The temperature ad relative humidity sesor is Sesirio SHT11 with accuracy of ο ± 0.4 C ad 3% o the temperature ad humidity, respectively. This sesor is coected to the MCU through the I C iterface. Other aalog sesors, such as soil moisture sesor ad light sesor ca be easily coected to the ADC iterface of the MCU. The SSN is powered by solar pael with 10.0 W. The RF module is ZigBee ad takes resposibility of trasferrig data i the etworks. A high power itegrated module which covers distace rage up to 1.5 Km is suitable for farmlad moitor. B. Wavelet-Based Sigal Compressio Algorithm Sigal compressio is a process that reduces the amout of data i order to reduce data trasmitted. Numerous of sigal compressio techiques have bee developed i the past few years, ad the wavelet trasform techiques have already achieved great success i the sigal compressio field. The wavelet trasform is a time-scale aalysis. The sigal is aalyzed at multiple frequecy rages with differet resolutios by decomposig the sigal ito a coarse approximatio ad detail iformatio. I order to save computatio eergy, we propose a waveletbased sigal compressio approach i WSN which does ot require computatio of certai high-pass coefficiets of the discrete wavelet trasform. This techique is called Skipped High Pass Sub-bads; SHPS [5]. Followig the same process give i [5], we use Mallat s pyramid algorithm [13] to implemet the DWT. The resultig wavelet coefficiets are the ecoded usig the SHPS techique. Fig. 6 illustrates the distributio of low-pass coefficiets (approximatio coefficiets) ad high-pass coefficiets (detail coefficiets) after applyig 1-level Haar wavelet trasform to the temperature sigal. The low-pass coefficiets represet approximatio of the origial sigal whereas the high-pass coefficiets represet detail iformatio of the origial sigal. We otice that the high-pass coefficiets are geerally small. Based o the umerical distributio, we ca estimate the highpass coefficiets to be zeros ad hece avoid computig them. The recostructio sigal without usig the high-pass coefficiets is show i Fig. 6 (d).the computatioal blocks i these realizatios are implemeted i the Simulik model as show i Fig. 7. Degree Celsius Fig. 4 Diagram of a wireless mesh etwork Mesh etwork Fig. 5 Implemetatio of a wireless mesh etwork Origial sigal (a) Number of coefficiets Low-pass coefficiets values (b) Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

5 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, Recostructio sigal without High-pass coefficiets Number of coefficiets Degree Celsius (c) (d) Fig. 6 (a) Origial sigal (b) umerical distributio of low-pass coefficiets (c) umerical distributio of high-pass coefficiets ad (d) recostructed sigal without usig high-pass coefficiets Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 IV. EXPERIMENTAL RESULTS AND DISCUSSIONS I order to evaluate the performace of the sigal compressio algorithm based o discretee wavelet trasform, we use MATLAB as simulatio tool. To demostrate the effectiveess of our proposed method, a series of experimets have bee coducted. The compressio ratio ( CR ), percetage root mea square differece ( PRD) ad quality factor ( QF ) are used as performace measures to quatify the differece betwee the origial sigal ad the processed sigal[15]. The compressio performace of the proposed algorithm is evaluated i terms of the CR which is defied as the ratio of the amout of ucompressed data size to the amout of compressed data size. The CR is defied as; Fig. 7 Simulik model I WSN sigal compressio, quality of the recostructed sigal is a importat issue. I this paper, the PRD is usually used for quality criterio. The PRD is defied as; N ( x[ ] xˆ [ ]) = PRD x [ ] 1 = N = 1 where x[] ad x ˆ[ ] deote data, respectively, ad N is the umber of samples withi oe data segmet. Fially, QF is defied as; 100% (13) the origial ad recostructed CR QF = (14) PRD Ucompressio size CR = Compressio size (1) A. Result of Software Simulatios For performace evaluatio, four kids of time series data, amely ambiet temperature, light itesity, relative humidity Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

6 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 ad the moisture cotet of soil are selected as the experimetal data series, with differet data fluctuatio characteristic. They are collected by four mesh odes at 5 miutes itervals from a farm at Suraaree Uiversity of Techology. The legth of each data series takes 88 poits ad they are show i Fig. 8. The choice of wavelet is importat i decompositio process ad the selected wavelet should be simple ad feasible for implemetatio i a resource-limited MCU. I additio, the fudametal of choosig low-frequecy coefficiets is to select them i a appropriate decompositio level. As a result, we ivestigate the choice of the most appropriate mother wavelet for SHPS sigal compressio techique. We have tested 3 wavelets: Haar, Daubechies s type 4 (DB4) ad Daubechies type 8 (DB8). The SHPS techique has bee carried out usig sigle level, -level ad 3-level wavelet decompositio. I this sectio, we used time series data, ad measured the CR, PRD ad QF of the compressed sigal. The results are preseted i Fig. 9. It ca be see from Fig. 9 that the choice of mother wavelet ad decompositio level have effect o the CR, PRD ad QF of the compressed sigal. Degree Celsius Humidity(%) Temperature (a) Humidity (c) Fig. 8 Origial sigals used i simulatios (a) temperature, (b) light, (c) humidity ad (d) moisture The experimets also demostrate that the SHPS techique allows us to choose sigal quality of wireless service by chagig the umber of wavelet decompositio levels. Thus, there will be a tradeoff betwee the sigal quality ad computatio complexity because the more levels of wavelet decompositio, the algorithm performs, the more computatioal time it takes. I this work, Haar wavelet at - level decompositio has bee chose for this implemetatio. Lux Moisture(%) 7 x 104 Light (b) Moisture (d) CR CR_Haar CR_DB4 CR_DB Level (a) QF CR QF_Haar QF_DB4 QF_DB8 (c) Fig. 9 (a) CR (b) PRD ad (c) QF as a fuctio of the mother wavelet ad decompositio level B. Result of Hardware Implemetatio QF I this experimet, we demostrate the efficiecy of this scheme uder real-world applicatios with limited umber of data samples. The four SSNs were deployed i a farmlad eviromet, all withi sigle-hop rage of the receiver device. The compressio ad trasmissio is doe automatically at the SSN base statio. After compressig, the compressed data is trasmitted to the web server. However, the trasmissio to the web server ca be wired or wireless based o the applicatio ad ecessity. The compariso of the origial sigal ad the recovery sigal of the temperature ad light itesity at the compressio ratio 1.7 are show i Fig. 11 ad Fig. 1, respectively. It is obviously show that the compressed data ca be recostructed effectively. Although some of the recostructed data has some differece from the origial data, there is almost o ifluece o the data series ad the total treds of data series are well illustrated. Therefore, the algorithm proposed i this paper ca be applied to compress the collected data ad reduce the data commuicatio as well as the eergy cosumptio of the SSN so that the lifetime of SSN ca be exteded. V. CONCLUSIONS Level This paper proposed a efficiet data compressio strategy exploitig the multi-resolutio characteristic of the wavelet trasform. We have developed a sesor ode called Smart Sesor Node; SSN. The mai goals of SSN desig are lightweight, miimal power cosumptio, modular desig ad robust circuitry. The SSN is made up of four basic compoets which are a sesig uit, a processig uit, a trasceiver uit ad a power uit. FiO Std evaluatio board is chose as the mai cotroller of the SSN for its low costs ad PRD PRD_Haar PRD_DB4 PRD_DB Level PRD (b) Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

7 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 high performace. The software codig of the implemetatio was doe usig Simulik model ad MATLAB programmig laguage. The experimetal results show that the proposed compressio algorithm yields recover sigal with good quality. Further research ca be cocetrated o the developmet of the optimizatio techique of compressio algorithm by usig the artificial itelliget techiques. Fig. 10 SSN base statio (Coordiator) istallatio i SUT farm Degree Celsius Fig. 11 Compariso of the origial sigal ad the recovery sigal of temperature with N = 64 Light(LUX) Temperature Light Origial sigal Recovery sigal Origial sigal Recovery sigal Fig. 1 Compariso of the origial sigal ad the recovery sigal of light itesity with N = 64 ACKNOWLEDGMENT This work was supported by a grat from SUT Research ad Developmet Fud, Suraaree Uiversity of Techology, Nakho Ratchasima, Thailad. REFERENCES [1] N. Watthaawisuth, N. Togrod, T. Kerdcharoe ad A.Tuatraot, Real-Time Moitorig of GPS-Trackig Tractor Based o ZigBee Multi-Hop Mesh Network, I Proc. the Electrical Egieerig/Electroics Computer Telecommuicatios ad Iformatio Techology, Vol. 1, pp , 010. [] N. Kimura ad S. Latifi, A Survey o Data Compressio i Wireless Sesor Networks," I Proc. the Iformatio Techology: Codig ad Computig, Vol., pp:8 13, 005. [3] A. Kulakov ad D. Davcev, Itelliget Data Acquisitio ad Processig Usig Wavelet Neural Networks, I Proc. IEEE Workshop o Itelliget Data Acquisitio ad Advaced Computig Systems: Techology ad Applicatios, Vol. 1, pp , 005. [4] A. Goh, S. Craciu, S. Rao, D. Cheey, K. Gugel, J. C. Sachez, J. C. Pricipe, Wireless Trasmissio of Neuroal Recordigs Usig a Portable Real-Time Discrimiatio/Compressio Algorithm," I Proc. the th Aual Iteratioal Coferece o Egieerig i Medicie ad Biology Society, pp: , 008. [5] M. Nasri, A. Helali, H. Sghaier, ad H. Maaref, Eergy-Efficiet Wavelet Image Compressio i Wireless Sesor Network," I Proc. the Commuicatio i Wireless Eviromets ad Ubiquitous Systems: New Challeges (ICWUS), pp. 1 7, 010. [6] N. Kimura ad S. Latifi, A survey o data compressio i wireless sesor etworks," I Proc. the Iformatio Techology: Codig ad Computig, Vol., pp:8 13, 005. [7] E. Chichi, H. Guyeet ad J. Friedt, K-RLE : A New Data Compressio Algorithm for Wireless Sesor Network, I Proc. the 009 Third Iteratioal Coferece o Sesor Techologies ad Applicatios, Vol. 1, pp , 009. [8] M. Nasri, A. Helali, H. Sghaier ad H. Maaref, Adaptive image trasfer for wireless sesor etworks (WSNs) I Proc. 010 Iteratioal Coferece o Desig & Techology of Itegrated Systems i Naoscale Era, Vol. 1, pp:1 6, 010. [9] E. Mahas, G. Brate, R. Souza ad M. Pellez, Eergy-Efficiet Cooperative Image Trasmissio Over Wireless Sesor Networks, I Proc. the 01 IEEE Wireless Commuicatios ad Networkig Coferece : Mobile ad Wireless Networks, Vol., pp , 01. [10] N.Rajput, N.Gadhi ad L. Saxea, Wireless Sesor Networks: Apple farmig i Norther Idia, I Proc. 01 Fourth Iteratioal Coferece o Computatioal Itelligece ad Commuicatio Networks, Vol. 1, pp. 18-1, 01. [11] M. Kohvakka, M. Kuorilehto, M. Häikäie ad T. D. Hämäläie, Performace Aalysis of IEEE ad ZigBee for Large-Scale Wireless Sesor Network Applicatios I. Proc. the 3rd ACM iteratioal workshop o Performace evaluatio of wireless ad hoc, sesor ad ubiquitous etworks, Vol. 1, pp.45-48, 006. [1] R. V. Kulkari ad G. K. Veayagamoorthy, Computatioal Itelligece i Wireless Sesor Networks: A Survey, IEEE Commuicatios Surveys & Tutorials, Vol. 13, No. 1, pp , 011. [13] S.G. Mallat, A Theory for Multiresolutio Sigal Decompositio: The Wavelet Represetatio, IEEE Tras. Patter Aal. Mach. Itell. Vol. 11, pp , [14] FiO Std evaluatio board web site, [15] B. Arviti, C. Naforita, I. Alexadru ad M. Costache ECG Sigal Compressio Usig Wavelets.Prelimiary Results, I. Proc th Iteratioal Symposium o Sigals, Circuits ad Systems, Vol. 1, pp. 1-4, 011. P. Kumsawat was bor i Maehogso, Thailad, i He received the B.Eg. degree i electrical egieerig from the Royal Thai Air Force Academy, Bagkok, Thailad, i 1994, the M.Eg. degree i electrical egieerig from Kasetsart Uiversity, Bagkok, Thailad, i 1997 ad Ph.D. degree i electrical egieerig from Suraaree Uiversity of Techology, Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514

8 Iteratioal Joural of Electrical, Computer, Eergetic, Electroic ad Commuicatio Egieerig Vol:7, No:5, 013 Nakho Ratchasima, Thailad, i 006. Sice 1999, he has bee with the Istitute of Egieerig, Suraaree Uiversity of Techology, Nakho Ratchasima, Thailad, where he is curretly a Assistat Professor i the School of Telecommuicatio Egieerig. His research iterests iclude digital sigal processig, image processig, wavelet trasform, multiwavelet trasform, artificial Itelligece, ad embedded systems. N. Pimpru was bor i Nakho Ratchasima, Thailad, i She received the B.Eg. degree i telecommuicatio egieerig from Suraaree Uiversity of Techology, Nakho Ratchasima, Thailad, i 011. She is curretly workig toward the M.Eg. degree i the School of Telecommuicatio Egieerig, Istitute of Egieerig, Suraaree Uiversity of Techology, Nakho Ratchasima, Thailad. Her research iterests iclude sigal processig, wireless sesor etworks ad artificial itelligece. Iteratioal Sciece Idex, Electroics ad Commuicatio Egieerig Vol:7, No:5, 013 waset.org/publicatio/514 K. Attakitmogcol was bor i Satu, Thailad, i 197. He received the B.Eg. degree i electroics egieerig from Kig Mogkut s Istitute of Techology, Ladkrabag, Bagkok, Thailad, i 1994 ad the M.S. ad Ph.D. degrees, both i electrical egieerig, from Vaderbilt Uiversity, Nashville, TN, i 1996 ad 1999, respectively. Sice 1999, he has bee with the Istitute of Egieerig, Suraaree Uiversity of Techology, Nakho Ratchasima, Thailad, where he is curretly a Associate Professor i the School of Electrical Egieerig. His research iterests iclude digital sigal processig, image processig, wavelet trasform, ad multiwavelet trasform. A. Srikaew was bor i Ubol Ratchathai, Thailad, i 197. He received the B.Eg. degree from Kig Mogkut s Istitute of Techology Ladkrabag, Bagkok, Thailad, i 1994 ad the M.S. ad Ph.D. degrees, both i electrical egieerig, from Vaderbilt Uiversity, Nashville, TN, i 1997 ad 000, respectively. Sice 000, he has bee with the School of Electrical Egieerig, Istitute of Egieerig, Suraaree Uiversity of Techology, Nakho Ratchasima, Thailad, where he is curretly a Associate Professor. His mai research iterests are i the area of computer ad robot visio, image processig, eural etworks, artificial Itelligece, ad itelliget systems. Iteratioal Scholarly ad Scietific Research & Iovatio 7(5) scholar.waset.org/1999.5/514