Online Signature Verification Using Angular Transformation for e-commerce Services

Transcription

1 Wold Academy of Scence, Engneeng and Technology Intenatonal Jounal of Electcal and Compute Engneeng Vol:4, No:1, 21 Onlne Sgnatue Vefcaton Usng Angula Tansfomaton fo e-commece Sevces Peeapong Uthansakul and Monthppa Uthansakul Intenatonal Scence Index, Electcal and Compute Engneeng Vol:4, No:1, 21 waset.og/publcaton/2291 Abstact The apd gowth of e-commece sevces s sgnfcantly obseved n the past decade. Howeve, the method to vefy the authentcated uses stll wdely depends on numec appoaches. A new seach on othe vefcaton methods sutable fo onlne e-commece s an nteestng ssue. In ths pape, a new onlne sgnatue-vefcaton method usng angula tansfomaton s pesented. Delay shfts exstng n onlne sgnatues ae estmated by the estmaton method elyng on angle epesentaton. In the poposed sgnatue-vefcaton algothm, all components of nput sgnatue ae extacted by consdeng the dscontnuous beak ponts on the steam of angula values. Then the estmated delay shft s captued by compang wth the selected efeence sgnatue and the eo matchng can be computed as a man featue used fo vefyng pocess. The theshold offsets ae calculated by two types of eo chaactestcs of the sgnatue vefcaton poblem, False Rejecton Rate (FRR) and False Acceptance Rate (FAR). The level of these two eo ates depends on the decson theshold chosen whose value s such as to ealze the Equal Eo Rate (EER; FAR = FRR). The expemental esults show that though the smple pogammng, employed on Intenet fo demonstatng e-commece sevces, the poposed method can povde 95.39% coect vefcatons and 7% bette than DP matchng based sgnatue-vefcaton method. In addton, the sgnatue vefcaton wth extactng components povdes moe elable esults than usng a whole decson makng. Keywods Onlne sgnatue vefcaton, e-commece sevces, Angula tansfomaton. I. INTRODUCTION ECENTLY, the gowth of puchasng mechandses va Relectonc onlne sevces so-called e-commece has been apdly dven by huge use demand. Howeve, the technology of use authozaton s stll based on numec o alphabet methods such as cedt cads, use name and passwod. These methods ae smple to mplement but along wth ease of fauds. Thee s an nceasng nteest n elable dentty vefcaton. Seveal bometc featues have been studed and poved useful, ncludng sgnatue, fngepnt, face, speech, s, and etna patten [1]-[2]. These bometc featues have advantages ove conventonal keys o pesonal dentfcaton numbes (PINs) and passwods n that they ae fee fom caage and memozaton poblems. Howeve, because most bologcal chaactestcs ae unchangeable, a moe seous The authos ae wth the School of Telecommuncaton Engneeng, Suanaee Unvesty of Technology, Nakhon Ratchasma, Thaland 3 (coespondng autho phone: ; fax: ; e-mal: uthansakul@sut.ac.th). Ths wok s fnancally suppoted by Suanaee Unvesty of Technoogy. poblem ases when they ae duplcated. Hence, one wll hestate to use the dsclosed bologcal featues. The onlne sgnatue s moe obust to the copy poblem than othe bologcal featues n that t has dynamc chaactestcs n addton to the mophologcal chaactestcs whle the othes geneally povde only the mophologcal chaactestcs [3]- [5]. As one can expect, t s vey dffcult to constuct a smlaly shaped sgnatue that also contans dynamc featues smultaneously. Moeove, one can change hs o he sgnatue ntentonally n case of secuty leakage. Because of the dynamc chaactestcs of onlne sgnatues, dentcal pats of sgnatues appea at dffeent tmes. Then thee exst delay shfts n sgnatue. In the paametc appoaches, they used global featues that ae not geatly affected by the nonlnea delay shfts [6]. Usng only global featues has the advantage of beng vey fast, but the eo ates ae geneally hgh. The functonal appoaches epesent the sgnatue sgnal as a functon of tme and compae the smlaty by accumulatng the dffeence between two functons n tme. The delay elatonshp between two functons s establshed by the dynamc pogammng (DP) matchng method [7] o hdden Makov Models (HMMs) [8]-[11]. In the DP matchng methods, they ted to fnd the exact matchng ponts between sgnatues by means of seachng whch esulted n too much computaton. In the HMM methods, much effot was spent on selectng the type of Makov models, and much computaton powe was used n enollment pocedues fo tanng the models. Recently, to get hghe secuty, hybd appoaches that use moe than one knd of featue have been eseached [12]. Whle they acque vey low eo ates, the pocessng tme seems to ncease. In ths pape, we popose a new method fo extactng sgnatue components by consdeng angle epesentaton n tme. Ths method splts a global delay shfts nto many delay shfts fo each component. Theefoe, t educes eo ate acqustons when combnng vefcaton esults of all components. The othe advantage of angula tansfomaton s that t educes the data sze of tansmsson on Intenet connecton because only angle chaactestcs ae sent fom clent to seve nstead of both vetcal and hozontal chaactestcs. In the poposed method, the estmated delay s captued by compang wth the selected efeence sgnatue and the eo matchng can be computed as a man featue used fo vefyng pocess. The theshold offsets ae calculated by two types of eo chaactestcs of the sgnatue vefcaton Intenatonal Scholaly and Scentfc Reseach & Innovaton 4(1) schola.waset.og/ /2291

2 Wold Academy of Scence, Engneeng and Technology Intenatonal Jounal of Electcal and Compute Engneeng Vol:4, No:1, Samplng tme Fg. 1 Demonstaton of angula tansfomaton Intenatonal Scence Index, Electcal and Compute Engneeng Vol:4, No:1, 21 waset.og/publcaton/2291 x-axs y-axs Samplng tme Fg. 2 Featues of sgnatue n (a) hozontal axs (b) vetcal axs (c) angle epesentaton poblem, False Rejecton Rate (FRR) and False Acceptance Rate (FAR). The level of these two eo ates depends on the decson theshold chosen whose value s such as to ealze the Equal Eo Rate (EER; FAR = FRR). All expements ae tested though exstng Intenet connecton. Own developed pogammng based Java Sevlet s employed. In patcula, the contbutons of ths pape ae ) the concept of usng angula tansfomaton fo onlne sgnatue vefcaton ) the vefcaton pocess usng many decdon makng fom extacted components and ) ts pefomances though eal-tme Intenet applcatons n compang wth othe methods. The emande of pape s oganzed as follows. In the followng secton, we descbe the onlne sgnatue pocessng ncludng angula tansfomaton and eo-delay estmatons. Secton III povdes the detals of pogammng fo e-commece sevces and then the vefcaton pocedue though such a pogam s descbed n Secton IV. Afte showng the expemental esults n Secton V, we conclude n Secton VI. Fg. 3 Example of extactng components fom angle epesentaton II. ONLINE SIGNATURE PROCESSING A. Angula Tansfomaton In geneal, the handwtng sgnatue ae captued and epesented n 2 dmensons, hozontal and vetcal axes. Fo angula tansfomaton, the doman consdeed nstead of both axes s the angle doman whch s calculated by slope fom one captued pont to anothe. The example of angula tansfomaton s shown n Fg. 1. Fg. 2 pesents the value of sgnatue n tem of hozontal, vetcal and angula values n tme. Fo poposed technque, only angula values of sgnatue ae tansmtted fom clent to seve. Hence, t saves some oveheads due to educng the sze of sgnatue featue. As seen n Fg. 2(c), the plot of angula epesentaton has many dscontnuous ponts. These postons ndcate two meanngs, shap tp and cossng p. Fo shap tp, t s the poston when stoke of sgnatue s stated. The othe dscontnuous poton s a cossng p because t s occued when angula value s changng between opposte sgn at 18 degee. Both meanngs ae helpful fo angula tansfomaton to extact sgnatue nto many components. The example of extacton s llustated n Fg. 3 and expessed n (1). M S( t) S ( t) (1) 1 whee M s the total components of sgnatue, S(t) s the angula epesentaton of handwtng sgnatue and S (t) s the th components of S(t) whch s defned by S( t) ts t te S ( t) (2) elsewhee whee ts s the statng tme of th component and te s the endng tme of th comoponent. B. Delay and Eo Estmatons The angula tansfomaton has been done n the clent sde and sent to seve. Afte extactng all components by seve, Intenatonal Scholaly and Scentfc Reseach & Innovaton 4(1) schola.waset.og/ /2291

3 Wold Academy of Scence, Engneeng and Technology Intenatonal Jounal of Electcal and Compute Engneeng Vol:4, No:1, Refeence sgnatue Delay sgnatue 45 4 Measued Estmated Delay tme t d slope = d Samplng tme Refeence tme t Intenatonal Scence Index, Electcal and Compute Engneeng Vol:4, No:1, 21 waset.og/publcaton/2291 Fg. 4 Compason between efeence and delay sgnatues of fst component shown n Fg. 3 the next pocess s to estmate delay shft and then calculate the eos. Fo onlne system, the tme usage of wtng sgnatue becomes the most mpotant nfomaton to be used fo vefyng pocess. Ths value s unquely dffeent fom offlne vefcaton because the offlne system cannot captue ths delay tme. In ths pape, the lnea appoxmaton s appled to estmate the delay. Fg. 4 shows the angula values of the fst component pesented n Fg. 3. The efeence sgnatue s the selected pototype used fo compang wth testng sgnatue. The method to choose efeence sgnatue wll be descbed n the next secton. As seen n Fg. 4, both sgnatue have a smla tend but speadng by dffeent tmes. Hence the delay shft d of th component can be defned as d S ( t) S ( d t) (3) whee (t) s the th component of efeence sgnatue and S d (t) S s the th component of delay sgnatue. To llustate the elaton between efeence and delay sgnatues, the tme ntevals of both sgnatues ae necessay to be nvestgated. Fg. 5 pesents tme ntevals of fst components gven n Fg. 4. It s clealy seen that the slope of Fg. 5 s appoxmately defned by delay shft d. Usng lnea appoxmaton shown n (4), then the delay shft can be easly estmated. d 2 d mn t d t (4) d t Afte estmatng the delay shft, the estmated sgnatue by compensatng the delay shft can be expessed as ~ d d S ( t) S ( t / d ) (5) Fg. 6 shows the compason between the fst components of efeence and estmated sgnatues deved by sgnatues pesented n Fg. 4. It s notced that thee s some eos Fg. 5 Appoxmatng the delay shft by slope of estmated lne Samplng tme between them whch summaton of squae eos can be expessed by ~ d e S ( t) S t ( t) Refeence sgnatue Estmated sgnatue Fg. 6 Compason between efeence and estmated sgnatues III. PROGRAMMING FOR E-COMMERCE SERVICES Ths pape ams to mplement the poposed technque fo e- Commece sevces. Theefoe, the platfom to suppot ths dea ae smulated by own developng pogam. In ths pape, the Java Sevlet s used to pogam both clent and seve ntefaces. Fg. 7 demonstates the connecton between clent and seve fo onlne vefcaton. Fo clent, the tablet pen mouse s attached to make a sgnatue wtng moe comfotable. Afte dgtal handwtng sgnatue s tansfomed nto angula domans, the clent pogam wll tansmt ths data to seve sde though nomal TCP/IP potocol n whch any ntenet bowse can be used. 2 (6) Intenatonal Scholaly and Scentfc Reseach & Innovaton 4(1) schola.waset.og/ /2291

4 Wold Academy of Scence, Engneeng and Technology Intenatonal Jounal of Electcal and Compute Engneeng Vol:4, No:1, 21 Fg.7 Connecton between seve and clent Intenatonal Scence Index, Electcal and Compute Engneeng Vol:4, No:1, 21 waset.og/publcaton/2291 Fo seve sde, all tanng sgnatues ae collected n the database fo compang wth the data tansmtted by clent. Only matched sgnatue wll be authozed nto the othe secton of web ste othewse the ejecton wll be tansmtted back to clent. Fg. 8 llustated a pogam sceen at clent sde opened by ntenet exploe. IV. VERIFICATION PROCEDURES A. Sgnatue Input Inteface Sgnatues ae nput by usng a commecal electonc pencl and a tablet. The appaatus used n the expements samples hozontal and vetcal postons 15 tmes pe second. The low-pass flteng and the sze and decton nomalzaton ae appled on the nput sgnatue sgnals. The featue pofle s extacted fom the pepocessed nput sgnal. All captued featues ae pefomed by Java pogammng. B. Angula Tansfomaton An angula tansfomaton s appled to the sgnatue featue pofle at clent sde. As descbed eale, the delay shft estmaton can be calculated by compang wth the efeence sgnatue. Howeve, only seve has these efeences, the pocess of estmatng delay shft s caed out at seve. Usng only ths angula featue, the component extacton s easly pefomed by consdeng the dscontnuous pont n tme. By extactng pocess, the seve can categoze sgnatues by numbe of components M. Ths can mpove the speed of coss check between smla sgnatue featues. In addton, t can ncease the qualty of vefcaton by M tme examnatons. C. Selecton of Refeence Sgnatue One authozed use has been equest to nput multple N sgnatues when egsteng. In ths pape only one efeence sgnatue s chosen fom those egsteed sgnatues. Howeve, the ctea to choose efeence sgnatue depends on the delay shft and eo estmatons. The aveage value of delay shft and eo estmatons when use each egsteed sgnatues as a efeenced sgnatue among those N sgnatues can be expessed by 1 N 1 2 l k d ( k) ag mn t d ( k, l) t (7) N 1 l d ( k, l) k t Aveage eo e (k) Fg. 8 Example of sceen shot at the clent sde to nput the sgnatue whee Aveage delay shft d (k) Fg. 9 Selectng efeence sgnatue 1 N 1 ~ 2 l k e ( k) S( t) S ( t) (8) N 1 k, l 1,..., N and k l l Fg. 9 pesents the example of aveage delay shft and aveage eos fom 1 egsteed sgnatues. The efeence sgnatue s selected fom the one closest to the cente. D. Selecton of Delay and Eo Thesholds Fo each authozed clent, vefcaton leanng s done only on egsteed sgnatues. The seve appoxmates, dung the vefcaton phase, the scoe of the sgnatue. Sgnatue vefcaton fo clent I was pefomed usng a decson t Intenatonal Scholaly and Scentfc Reseach & Innovaton 4(1) schola.waset.og/ /2291

5 Wold Academy of Scence, Engneeng and Technology Intenatonal Jounal of Electcal and Compute Engneeng Vol:4, No:1, 21 Pobablty TABLE II SIGNATURE VERIFICATION RESULTS FOR 1 CLIENTS FRR by egsteed sgnatues Clent FRR by tested sgnatues FAR by fogees Coect Offset FRR FAR Methods TABLE III COMPARISON WITH OTHER METHODS All tested Sgnatues FRR FAR PoposedFg. 11 Decson 2.37makng 2.24 pocess All Fogees FRR FAR Angula wthout extactng components DP matchng Basc matchng Intenatonal Scence Index, Electcal and Compute Engneeng Vol:4, No:1, 21 waset.og/publcaton/ Decson theshold Fg. 1 FAR and FRR cuves theshold. A decson s taken about whethe the supposed sgnatue of clent I s authentc o foged by compang the scoe to ts decson theshold. Two types of eo ae chaactestc of the sgnatue vefcaton poblem: Eo I o false ejecton ate (FRR), whch s when an authentc sgnatue s ejected, and Eo II o false acceptance ate (FAR), whch s when a fogey s accepted. The level of these two eo ates depends on the decson theshold chosen; we look fo a decson theshold whose value s such as to ealze the Equal Eo Rate EER (FAR = FRR). The decson theshold conssts of two tems: 1) the decson 4.41 theshold 4.23 obtaned at the pont whee the FRR cuve estmated wth the egsteed sgnatues used fo tanng and the FAR cuve estmated wth some fogees coss, and 2) an offset common to all clents because the FRR cuve obtaned wth the tanng sequences s based. To estmate the FAR cuve, 2.48wth 2.36 some fogees, ths pape uses, n the andom fogees expement, the othe tanng egsteed sgnatues and, n the smple fogees expement, we estmate the FAR cuve wth thee smple fogees dscaded fo test puposes The offset s obtaned as shown n Fg. 1. We estmate the FRR cuve wth egsteed sgnatues fo tanng, the FRR cuve wth sgnatues wote by those authentc clents fo testng, and the FAR cuve fo fogees fo 1 andomly selected clents of the database E. Decson Makng As descbed eale, t s easy to extact sgnatue nto many components usng angula consdeaton. Hence, thee ae many featues of sgnatue whch s able 97.8to make 1.12 a decson 1.8 sepaately. Fo fnal decson, the authentc clent s accepted when all components have to be passed the desed theshold. The pocess of decson makng s llustated n Fg. 11. V. EXPERIMENTS AND RESULTS The dataset used n the expement s composed of 3 sgnatues poduced by 1 clents wtng 2 egsteed/tested TABLE I SIGNATURE VERIFICATION RESULTS FOR NUMBER OF TRAINING SIGNATURES Numbe of Tanng All tested Sgnatues All Fogees Sgnatues FRR FAR FRR FAR sgnatues and 1 fogey sgnatues foged by 1 voluntees. The seve and clent ae on the same local aea netwok. The poposed sgnatue-vefcaton algothm was mplemented wth Java Sevlet comple. We compaed the pefomance of the poposed algothm wth a DP matchng-based functonal appoach [7] whch was developed by one of the authos. The DP matchng-based method selects multple pototypes by usng ISODATA clusteng algothm. A. Numbe of Tanng Sgnatues It s natual that the possblty of selectng an appopate efeence sgnatue nceases as the numbe of tanng samples nceases. In addton, because the decson theshold s computed fom FAR and FRR among the tested sgnatues, f the numbe of tanng samples s too small, selectng the decson bounday fo decdng authentc o foged sgnatues wll also be dffcult. Table I shows the vefcaton esults when nceasng numbe of tanng sgnatues. The esults fom all egsteed sgnatues ae bette than fogees. Ths s because the fogey sgnatues have moe vaety featues than egsteed clents. B. Vefcaton Results Table II pesents the vefcaton esults of 1 clents fo 1 tanng sgnatues. The poposed technque can offe aveage 95.39% coect vefcaton. It s also notced that the complex sgnatues povde low coect ate and hgh FRR/FAR. C. Compason wth Othe Methods To llustate the pefomance of ou poposed method, we compaed the esults wth othe appoaches whch ae the angula tansfomaton wthout extactng components, DP Intenatonal Scholaly and Scentfc Reseach & Innovaton 4(1) 21 1 schola.waset.og/ /2291

6 Wold Academy of Scence, Engneeng and Technology Intenatonal Jounal of Electcal and Compute Engneeng Vol:4, No:1, 21 Intenatonal Scence Index, Electcal and Compute Engneeng Vol:4, No:1, 21 waset.og/publcaton/2291 matchng [7] and basc matchng methods. Fo poposed method, the decson makng s done unde the concluson of all components extacted fom sgnatue. Howeve, t s nteestng to see whethe only one featue wthout extacton poduces the dffeent esults o not. Fo basc matchng method, we use fou combnatons of sgnatue s length, tme, sze and numbe of tunng cuve. Table III pesents the compason esults between poposed method and othes. The poposed method can povde 7% and 7.7% bette than DP matchng based sgnatue-vefcaton method fo all tested sgnatues and fogees espectvely. In addton, the sgnatue vefcaton wth extactng components offes moe elable vefcaton esults than usng a whole decson makng. Ths s because the use of component extacton can sceen dssmlates of othe sgnatues befoe makng a decson. As seen n Table III, t s obvous that the basc matchng appoach s the wost vefcaton but t needs, howeve, the least complexty fo mplementaton. VI. CONCLUSION In ths pape, we popose an onlne sgnatue vefcaton usng angula tansfomaton fo e-commece sevces. To moe effectvely avod the copy poblem, the angula featues of sgnatues n the fom of delay shfts ae examned. In addton, t s easy fo angula featue to extact a whole sgnatue nto many components so the decson makng has to be eled on ndvdual judgment of each component. Ths pocedue nceases the vefcaton coecton ate. By own developng seve-clent pogam, the vefcaton esults ae able to collect though Intenet connecton. Expemental esults show that the poposed method could compae moe accuate than othe appoaches. REFERENCES [1] M. Fahust, K. Cowley, and E. Sweeney, KAPPA Automatc Sgnatue Vefcaton: Sgnatue Vefcaton Publc Tals and Publc Suvey on Bometcs, Btsh Technology Goup, Tech. Rep., [2] C.-C. Hsu, L.-F. Chen, P.-C. Chang, and B.-S. Jeng, On-lne chnese sgnatue vefcaton based on mult-expet stategy, n Poc. 32nd Int. Canahan Conf. Secuty Technology, 1998, pp [3] R. Plamondon and G. Loette, Desgnng an automatc sgnatue vefe: Poblem defnton and system descpton, n Compute Pocessng of Handwtng, R. Plamondon and C. G. Leedham, Eds. Sngapoe: Wold Scentfc, 199, pp [4] S. H. Km, M. S. Pak, and J. Km, Applyng pesonalzed weghts to a featue set fo on-lne sgnatue vefcaton, n Poc. 3d Int. Conf. Document Analyss and Recognton,Monteal, QC, Canada, Aug. 1995, pp [5] Y. Sato and K. Kogue, On-lne sgnatue vefcaton based on shape, moton, and wtng pessue, n Poc. IEEE Int. Conf. Patten Recognton, vol. 2, 1982, pp [6] L. L. Lee, T. Bege, and E. Avcze, Relable on-lne human sgnatue vefcaton systems, IEEE Tans. Patten Anal. Machne Intell., vol. 18, no. 6, pp , Jun [7] J. R. Yu, S. H. Km, and J. Km, A class leanng method fo sgnatue vefcaton usng dynamc pogammng, J. Koea Inst. Telemat. Electon., vol. 32-B, no. 2, pp , [8] B. Wtz, Stoke-based tme wapng fo sgnatue vefcaton, n Poc. Int. Conf. Document Analyss and Recognton, vol. 1, 1995, pp [9] J. G. A. Dolfng, A compason of lgatue and contextual models fo hdden Makov models based on on-lne handwtng ecognton, n Poc. Int. Conf. Acoustcs, Speech, and Sgnal Pocessng, vol. 2, 1998, pp [1] M. Fuentes, S. Gaca-Salcett, and B. Dozz, On Lne Sgnatue Vefcaton: Fuson of a Hdden Makov Model and a Neual Netwok va a Suppot Vecto Machne, Poc. Eghth Int l Wokshop Fontes n Handwtng Recognton, 22, pp , Aug. 22. [11] J. G. A. Dolfng, Handwtng ecognton and vefcaton: A hdden Makov appoach, Ph.D. dssetaton, Technsche Unvestet Endhoven, Endhoven, The Nethelands, [12] N.-J. Cheng, K. Lu, K.-C. Cheng, C.-C. Tseng, and B.-S. Jeng, Onlne chnese sgnatue vefcaton usng votng scheme, n Poc. 31st Annu IEEE Int. Canahan Conf. Secuty Technology, 1997, pp [13] A. Kandel, Fuzzy Technques n Patten Recognton. New Yok: Wley, [14] M. Nadle and E. Smth, Patten Recognton Engneeng. New Yok: Wley, 1993, pp Peeapong Uthansakul (M 7) eceved B.Eng and M.Eng degees fom Chulalongkon Unvesty, Thaland n 1996 and 1998, espectvely. In , he woked as Telecommuncaton Engnee wth Telephone Oganzaton of Thaland (TOT) and then he has joned Suanaee Unvesty of Technology snce 2. Dung 23-27, he studed PhD at Unvesty of Queensland, Austala, n the aea of weless communcatons especally MIMO technology. He cuently woks as Assstant Pofesso n school of Telecommuncaton Engneeng, Faculty of Engneeng, Suanaee Unvesty of Technology, Thaland. He wote 1 book enttled Adaptve MIMO Systems: Exploatons fo Indoo Weless Communcatons (also avalable on amazon.com) and he has publshed moe than 6 efeee jounal and confeence papes. Hs cuent eseach nteests nclude MIMO, OFDM, WMAX and Weless Mesh Netwok. D. Uthansakul eceved 25 Best Student Pesentaton Pze wnne at the 9th Austalan Symposum on Antennas, Sydney, Febuay 25, Austala and 24 Young Scentst Tavel Gant wnne at the 24 Intenatonal Symposum on Antenna and Popagaton, August 24, Japan. Monthppa Uthansakul (M 7) eceved B.Eng degee fom Suanaee Unvesty of Technology, Thaland, n 1997 and M.Eng degees fom Chulalongkon Unvesty, Thaland n She has joned Suanaee Unvesty of Technology snce Dung 23-27, she studed PhD at Unvesty of Queensland, Austala, n the aea of smat antenna especally wdeband beamfomng. She cuently woks as Assstant Pofesso n school of Telecommuncaton Engneeng, Faculty of Engneeng, Suanaee Unvesty of Technology, Thaland. She wote 1 book chapte enttled Wdeband smat antenna avodng tapped-delay lnes and fltes n Handbook on Advancements n Smat Antenna Technologes fo Weless Netwoks, Idea Goup Publshng, USA, 28 and she has publshed moe than 5 efeee jounal and confeence papes. He cuent eseach nteests nclude antenna aay pocessng, compact swtched beam antenna and body communcatons. D. Uthansakul eceved Young Scentst Contest 2nd Pze at 16th Intenatonal Confeence on Mcowaves, Rada and Weless Communcatons, Kakow, Poland, May 26. Intenatonal Scholaly and Scentfc Reseach & Innovaton 4(1) schola.waset.og/ /2291