3rd Iteratioal Coferece o Mechatroics ad Iformatio Techology (ICMIT 206) A Novel Iterferece Suppressio Algorithm Based o Irregular Wavelet Pacet Trasform i DSSS Satellite Commuicatio System Yaghui Tog, Fagju Liu, Daoxig Guo ad Heg Wag PLA Uiversity of Sciece ad Techology, Najig, Jiagsu, Chia email:024276863@qq.com Keywords: satellite commuicatios, DSSS, wavelet pacet trasform, Optimum wavelet base, iterferece suppressio Abstract: I recet year, the Direct Sequece Spread Spectrum (DSSS) techology has bee itroduced as a attractive approach to guaratee ati-iterferece capability i satellite commuicatio system. However, due to the limitatio of the DSSS, the system performace has bee greatly deteriorated while a strog iterferece is out of the tolerace of the system. Meawhile, what the wavelet pacet trasforms is ideally suitable for the iterferece detectio ad suppressio of DSSS satellite commuicatio system by exploitig its excellet local time-frequecy domai aalysis capability. I this paper, a ovel algorithm has bee put forward to achievig positio ad iterferece suppressio through combig the sub-bad power ratio with miimum power thresholds which are based o the aalysis o the optimal wavelet pacet decompositio. Simulatio result shows that the proposed iterferece suppressio algorithm sigificatly improves the ati-jammig capability of DSSS satellite commuicatio system compared with the traditioal FFT algorithm. Itroductio With the advet of the iformatio age, the requiremets of accessig to iformatio are icreasigly urget. Due to the ati-iterferece capacities of direct sequece spread spectrum (DSSS) systems, it has bee widely used i both the military ad civil commuicatios. However, the ati-iterferece capability of DSSS system is limited by the processig gai. The performace of the system largely degrades whe the iterferece power is greater tha the jammig margi. It is of great sigificat to employ sigal processig techiques to improve the ati-jammig performace of DSSS system. Curretly, the sigal processig techiques is used i the spread spectrum system which is iclude time domai processig techiques [, 2] ad trasform domai processig structures [3, 4]. Time domai processig techiques ca elimiate the arrowbad iterferece (NBI) completely because it estimates the iterferece exactly ad extract from the receive sigal to leave the iterferece free DSSS sigal. But it is ecessary for covergece time to reach the optimal solutio. So it is suitable for slow-altered iterferece. O the other had, the chagig iterferece ca be quicly traced by the trasform domai suppressio. The trasform domai processig structures adopt the methods lie the Fourier trasform, wavelet trasform ad wavelet pacet trasform ad to covert a time-domai sigal ito the trasform domai. It utilizes the differet features betwee the iterferece sigal ad the desired sigal i the frequecy domai to mae a distictio. It uses the relevat suppressio method to elimiate the iterferece sigals. Therefore, the trasform domai processig techiques are able to trac ad capture dyamic iterferece quicly ad adaptively. Wavelet pacet trasform is very suitable for iterferece detectio ad suppressio i the direct sequece spread spectrum systems because of its excellet local time ad frequecy domai aalysis capabilities. The difficulty of usig wavelet pacet to achieve iterferece suppressio is maily reflected i two aspects: The first oe is about how to carry out wavelet pacet decompositio o the received sigal. The secod oe is about how to locate iterferece o the basis of wavelet tree ad to choose appropriate algorithms for achievig iterferece suppressio. As is ow to all, the wavelet pacet decompositio usually icludes regular decompositio 206. The authors - Published by Atlatis Press 05
with uiform sub-bad ad optimal wavelet pacet decompositio with ueve sub-bad [5,6]. The commo iterferece locatig methods based o wavelet pacet iclude sub-bad power ratio ode positioig [7], adaptive threshold positioig [8] ad so o. A ew method for iterferece suppressio has bee put forward based o the optimum wavelet pacet decompositio. Firstly, locate the iterferece by usig sub-bad power ratio so that sub-bad high-power iterferece ca be suppressed. The, mae further judgmets ad suppressio o the o-suppressed sub-bad by usig the miimum power threshold method i order to have a better elimiate residual iterferece. System model The iterferece suppressio ad oise suppressio model,which is built based o wavelet pacet, which is show i Fig.. The sigal from the groud receiver usually icludes spread spectrum sigal compoet, arrowbad iterferece ad oise compoets. The received sigal r ca be expressed as: ( ) r( ) = s( ) + J( ) + ( ) () I this formula, the s( ) deotes spread spectrum commuicatio sigal usig the BPSK modulatio, J( ) presets the arrow-bad iterferece sigals, ( ) deotes the additive 2 white Gaussia oise sigal with a mea zero ad the variace σ. Thus, the sigal s( ) based o direct sequece spread spectrum system ca be expressed as: s = Ps PN cos ω + φ (2) ( ) 0( ) ( ) [ 0 ] I this equatio above, P represets power level of the spread spectrum sigal, s0 ( ) is the biary iformatio bits, PN( ) is the spreadig sequece, ω 0 is carrier frequecy, φ is the phase. Satellite Iformatio traspoder iput Upli iterferece Wavelet pacet trasform Orietatio ad iterferece suppressi o Trasform domai si gal processi g Wavelet pacet Atitrasform Iformatio output Fig. Schematic of DSSS iterferece suppressio satellite commuicatio based o wavelet pacet trasform Wavelet pacet aalysis basis of sigal Wavelet pacet trasform has a excellet time-frequecy localizatio features ad multi-resolutio aalysis ability. Whe the iterferece chages i real-time, the iterferece will be located quicly ad efficietly i a limited sub-bad ad the it will be elimiated through the relevat suppressio algorithm. The trasform process of Wavelet pacet is defied usig the followig sequece of fuctios with recursio: (3) () = 2 ( ) ( 2 ) () = 2 ( ) ( 2 ) U t h U t U t g U t 2 2+ Z Z I this formula, U0 ( t ) is the scalig fuctio of φ ( t ), U ( ) ψ ( t ), { U ( t), Z} is ow as Wavelet Pacet Group of U0 ( t ), { h( ), Z} t is the mother wavelet of ad 06
{ g( ) ( ) h( L, ) Z} = respectively represets a low-pass filter coefficiet group ad high-pass filter set of coefficiets of quadrature mirror filters QMF with supportig legth L, ad satisfy the followig coditio: h( 2a) h( 2b) = δab, Z (4) h( ) = 2 Z Wavelet pacet eeds to be discreted for practical applicatios. A followig recursive discrete wavelet pacet trasform was give by C.K.Chui[9]: 2 Sl+ () i = hi ( 2) Sl ( ) z 2+ Sl+ () i = g( i 2) Sl ( ) z The correspodig iverse discrete wavelet pacet trasform is as follows: (6) () = ( 2 ) ( ) + ( 2 ) ( ) S i hi S g i S 2 2 2 l l+ l+ z z I this formula, l represets the correspodig layers of wavelet pacet decompositio, idicates the lateral odes positio of the correspodig level; S represets the decompositio sequece of ode at the layer of l. The received sigal ca be separated to a uiform or o-uiform spectral sub-bad by maig use of wavelet pacet trasform. Wavelet pacet decompositio of the DS sigal The optimum wavelet pacet base is foud out by the rule of eergy compact i this paper, the specific steps are as follows: () Mae use of M -ary wavelet pacet to decompose the received spread spectrum sigal to gai a M -ary rules wavelet pacet decompositio tree. The umber of odes cotaied i every layer ( L ) is M. (2) Each ode uses a stadard eergy compact G to aalyze: M M 2 2 G = σ / σ (7) = 2 σ is the coefficiet variace of ode N, 2 σ is the coefficiet variace of layer i ( i L ) of the child ode N whose umber is M ad paret ode is N of the layer i+. The threshold is Th_ ECM, if the stadard G satisfies the coditio G > Th_ ECM, it idicates that the eergy of child ode N is ueve so it eed eep o decomposig; if the stadard G satisfies the coditio G Th_ ECM, it idicates that the eergy of child ode N is eve ad the most parts of its frequecy bad are composed of spread spectrum sigal elemet or iterferece elemet, so i this case the ode will o loger mae further decompositio. It is worth oticig that it eeds to mae iterferece judgmet of the leaf odes that are out of breaig dow at the time of iterferece positio ad suppressio to avoid the ifluece of iterferece compoets o system performace. (3) If the ode eeds further decompositio, doig wavelet pacet trasform agai is ecessary to have the child ode N of the layer i+ 2, ad the repeat step () i a similar operatio util the l (5) 07
decompositio achieves the give maximum umber of scale L (At this time it correspods to the maximum frequecy resolutio), the odes stop decomposig, ad we will get a optimum wavelet pacet base, the diagram show i Fig.2 is a excellet example. (0,0) (,0) (,) (2,0) (2,) (3,0) (3,) 0 π 4π (4,0) (4,) (4,2) (4,3) 6 6 Fig.2 Diagram of optimal wavelet pacet decompositio tree (decompositio level of 4) ad the correspodig rage badwidth decompositio 8π 6 π ω The iterferece positioig ad suppressio algorithm o the basis of wavelet pacet trasform. The steps preset the iterferece suppressio method of combiig the sub-bad power ratio with the miimum threshold i this article are as follows: () Read the leaf ode leaves of optimal wavelet pacet tree, loo for double leaves odes which have the same paret ode, ad process those double leaves odes with the sub-bad power ratio iterferece positioig suppressio. The use the miimum threshold to judge the existece of the residual iterferece, if there is iterferece, the the coefficiet of this leave is set as zero directly; (2) Search for sigle leaf odes which have differet paret odes. Accordig to the process of optimal wavelet pacet decompositio, there may be iterferece left o the decomposed sigle leaf odes. I order to achieve better ihibitory effect, read coefficiets of all leaf odes ad calculate the variace firstly, the mae use of the media i the variace collectio to judge those leaves. If the variace of the leaves is greater tha the media, it idicates that iterferece exists i the sigle leaf ode, the the coefficiet of this leave should be set as zero directly, otherwise, there is o iterferece i this sigle leaf ode, which avoids the ifluece whe the sigle leaf ode is affected by iterferece. At the same time, it also esures that most of the useful sigals are udamaged; (3) Trasform the processed wavelet pacet tree iversely by adoptig M-ary wavelet pacet ad obtai the spread spectrum sigal through iterferece suppressio, the mae despreadig, demodulatio ad other subsequet processig. The simulatio results ad aalysis I the simulatio process, the DS system uses a pseudo-radom sequece of legth 32 to spread, modulatio scheme is BPSK modulatio, wavelet pacet decompositio taes geeratio fuctio of db6 (Daubechies wavelets), M of M-Ary taes two, the chael of AWGN is tae. Because the 08
spreadig code legth is 32, the system has a certai ati-iterferece tolerace, i order to reflect the performace of the algorithm for iterferece suppressio well, jammig-to-sigal ratio (JSR) must be greater tha 5.05dB ( 0 log0 ( 32) 5.05dB) =, jammig-to-sigal ratio is from 20 to 50dB i simulatio process. The maximum decompositio level of wavelet pacet tree is 5,what is eeded to satisfy both the suppressio performace ad the complex of decompositio. The ormalized digital frequecies of two-toe iterferece to spreadig rate are 0.34 ad.57, phase i [ 0,2π ] is uiform. From the Fig.3, it ca be see that the iterferece for the DS system poses a serious deterioratio, it ca ot commuicate properly without suppressio, while the trasform domai processig structures ca effectively improve the performace of the system. Furthermore, the ovel iterferece suppressio algorithm based o irregular wavelet pacet trasform of combiig the sub-bad power ratio ad the miimum threshold value has the better ihibitory effect tha the traditioal FFT iterferece suppressio. From the Fig.4, i the same coditio that the sigal-to-oise ratio (SNR) is 8dB, whe SNR is low, the performace of the ovel iterferece suppressio algorithm based o irregular wavelet pacet trasform of combiig the sub-bad power ratio ad the miimum threshold value is close with the traditioal FFT iterferece suppressio. But as the SNR icreases, the ati-iterferece effect of the latter is sigificatly better tha the former. 0 0 0 0 0-0 - 0-2 P b 0-2 P b 0-3 No iterferece suppressio FFT iterferece suppressio A combied suppressio of the sub-bad power ratio ad the miimum threshold value Oly the presece of oise 0-3 0-4 No iterferece suppressio FFT iterferece suppressio A combied suppressio of the sub-bad power ratio ad the miimum threshold value Oly the presece of oise 0-4 20 25 30 35 40 45 50 JSR[dB] Fig.3 The BER performace compariso of various algorithms i differet jammig-to- sigal ratio ad two-toe iterferece (SNR = 8dB) 0-5 2 3 4 5 6 7 8 9 SNR[dB] Fig.4 The BER performace compariso of various algorithms i differet sigal-tooise ratio ad two-toe iterferece (JSR = 30dB) Coclusio A ovel iterferece suppressio algorithm that irregular wavelet pacets trasform by combiig the sub-bad power ratio ad the miimum threshold is proposed i this paper. Our 09
proposed algorithm limits the iterferece to a few umber of sub-bads to avoid damage to the useful sigal while thoroughly elimiates the residual iterferece with the assistace of sub-bad power ratio suppressio. Simulatio results reveals that iterferece are effectively located ad suppressed eve through uder the sceario of strog iterferece, which sigificatly ehace the ati-iterferece ability of DSSS satellite commuicatio system. Refereces []. K.C.Ho, Xiaoig Lu, Vadaa Metha. Adaptive Blid Narrowbad Iterferece Cacellatio for Multi-User Detectio[J]. IEEE Trasactios o Wireless Commuicatios, 2007,6(3):024-033. [2]. Rodrigo C.de Lamare, Marti Haardt, Raimudo Sampaio-Neto. Blid Adaptive Costraied Reduced-Ra Parameter Estimatio Based o Costat Modulus Desig for CDMA Iterferece Suppressio[J]. IEEE Tras o Sigal Processig, 2008, 56(6): 2470-2482. [3]. Chuhai Zhag, Liju Xue, Eryag Zhag. Narrow-bad iterferece suppressio i trasform domai based o adaptive multi-threshold algorithm [J]. Joural of Electroics & Iformatio Techology, 2006, 28(3):46-465. [4]. P.Azmi ad M.Nasiri-Keari. Geeralised Fourier trasform-domai techique for arrowbad iterferece suppressio i CDMA commuicatio system[j]. Electroics Letters,200,37(0):652-654. [5]. Shi Xiaju, Li Sa, Li Ruiliag. Sigal deoisig algorithm ad its applicatio based o best wavelet pacet basis[j]. Joural of Naval Aeroautical Egieerig Istitute, 2006,2(5):506-509. [6]. Jiag Yu, Xiao Hog, Teg Wei, Liu Xigpeg, Gao Hogyou, Yu Shaopeg. Wavelet-pacet-trasform for arrow-brad-iterferece suppressio i direct sequece spread spectrum commuicatio system[j]. Joural of Harbi Uiversity of Commerce,2008,24():98-00. [7]. Zhu Liwei, Jiag Piqu. Adaptive Threshold Based o Wavelet Pacet Trasform for Narrowbad Iterferece Suppressio[J]. Joural of Data Acquisitio ad Processig,203,28(6):843-847. [8].Emilia Pardo, Miguel A. Rodriguez-Heradez, Jua J.Perez-Solao. Narrowbad iterferece suppressio usig udecimated wavelet pacets i direct-sequece spread-spectrum receivers[j]. IEEE Trasactios o Sigal Processig, 2006,54(9):3648-3653. [9]. Weimi Yag, Guagguo Bi. Rejectio of arrowbad iterferece i DSSS systems based o adaptive wavelet pacet trasform[j]. Joural of Chia Istitute of Commuicatios,999, 20(7):69-75. 0