ࡱ> bdagY -UbjbjWW "==G ]22222,2X" cq]!!!!!!!$#%!U_cUU!i.. iiiU.  !@@@@..U!ii5 V!!@6hm! $I2j22Ka! SPARSE VECTOR LINEAR PREDICTION WITH NEAR-OPTIMAL MATRIX STRUCTURES Davorka Petrinovi, Davor Petrinovi Faculty of Electrical Engineering and Computing, Zagreb, Croatia Abstract Vector Linear Prediction (VLP) is frequently used in speech and image coding. This paper addresses a technique of reducing the complexity of VLP, named the sparse VLP (sVLP), by decreasing the number of nonzero elements of prediction matrices used for prediction. The pattern of zero and nonzero elements in a matrix, i.e. the matrix structure, is not restricted in the design procedure but is a result of the correlation properties of the input vector process. Mathematical formulations of several criteria for obtaining near-optimal matrix structures are given. The consequent decrease of the sVLP performance compared to the full predictor case can be kept as low as possible by re-optimizing the values of matrix non-zero elements for a resulting sparse structure. Effectiveness of the sVLP is illustrated on vector prediction of the Line Spectrum Frequencies (LSF) vectors and compared to the full predictor VLP. Keywords Vector Linear Prediction, sparse matrices, complexity reduction, LSF INTRODUCTION Vector process can be described as a process whose input samples are vectors. For a particular class of vector processes, each component xi(n), i=1,,k of the current vector x(n) can be estimated based on linear combination of all components of certain number of preceding vectors as in:  EMBED Equation.3 ( SEQ ( \* ARABIC 1)This technique, known as vector linear prediction (VLP) is frequently used for signal coding (e.g. Yong, 1988) by performing quantization on the prediction residual e(n),  EMBED Equation.3 ( SEQ ( \* ARABIC 2)i.e. the difference between the original and the predicted vector, instead of the original vector. The prediction gain that can be achieved by such predictive quantization is defined as:  EMBED Equation.3 ( SEQ ( \* ARABIC 3)where E denotes the statistical expectation and |||| denotes Euclidean norm. It depends on the degree of correlation between consecutive vectors of the frame. This correlation is modeled by one or more predictor matrices Am, depending on the prediction order M. For simplicity, only the first order of prediction (M=1) will be discussed in the paper. The optimal predictor A minimizing the prediction residual energy within the analysis frame (block of input vectors) can be found by solving normal equations as in (Chen, 1987), that results with:  EMBED Equation.2   EMBED Equation.3 ( SEQ ( \* ARABIC 4)where C={ci,,j} and  EMBED Equation.3 ={gi,,j} are covariance matrices calculated from the input vector process comprised of vectors x(0) to x(P). Although VLP can maximally exploit interframe as well as intraframe correlation by predicting each component of the current vector from all components of the previous vector, it is a reasonable assumption that for a certain class of vector processes not all components of the preceding vector contribute equally to prediction. This gives rise to an idea of simplifying the VLP by employing the predictor matrix A that is sparse, i.e. that models only the correlation between the most significant components of two consecutive vectors, while for other components the correlation is not modeled at all (A has zeros at those positions). Such procedure should not have a great impact on prediction gain, but can at the same time reduce the amount of computation in coding techniques based on VLP. The main problem of the proposed sparse VLP (sVLP) design technique is to determine which elements of predictor matrix can be zeroed, i.e. what is the suitable sparse matrix structure. For example, for vector processes that exhibit strong correlation between only a few neighboring components, a predictor with predefined multidiagonal structure is suitable (Petrinovi, 1999). The number of diagonals can be chosen as a design parameter, determining the total number of nonzero elements. In this paper, a more general predictor structures are discussed, that are unrestricted and depend on the actual correlation of the input vector process. Several techniques for obtaining near-optimal sparse predictor structures are investigated and illustrated. A design procedure for obtaining an optimal sparse predictor that maximizes the prediction gain for any structure will be described. Comparative simulation results for several proposed criteria are also given. SPARSE PREDICTORS In order to establish suitable sparse structures of a predictor that maximize the prediction gain for any chosen number of nonzero elements, it is necessary to determine the contribution of each predictor matrix element to the prediction gain. According to  REF _Ref479157793 \h (3), prediction gain is defined by the energy of the prediction residual that can be expressed as a function of predictor elements. If the elements of the predictor are zeroed one at a time and the change of residual energy is calculated for each structure, it would be possible to determine the element that results with the minimal change. This component of the predictor is then set to zero, residual energies are recalculated and the whole procedure is repeated until the desired number of nonzero elements is reached. In order to implement the above design procedure, the differential increase of the energy of the prediction residual due to zeroing any predictor element has to be established first. Let ei(n) denote the prediction error (residual) sequence of the ith vector component expressed as:  EMBED Equation.3 ( SEQ ( \* ARABIC 5)where  EMBED Equation.3  represents the ith row of matrix A. For the classical VLP based on a full predictor, the residual energy Efi of the ith component can be expressed as in:  EMBED Equation.3 ( SEQ ( \* ARABIC 6)E0i is the energy of the ith component of the input vector process, and  EMBED Equation.3  is the ith column of the covariance matrix  EMBED Equation.3 . If  EMBED Equation.3 is a row of an optimal predictor A found according to expression  REF _Ref478283845 \h (4), then  EMBED Equation.3  is satisfied for all  EMBED Equation.3  and  REF _Ref479145595 \h (6) can be simplified to:  EMBED Equation.3 ( SEQ ( \* ARABIC 7)The total residual energy is equal to the sum of component residual energies, Ef1 to Efk, and is minimum if all of the Efi are also minimum. It is obvious from  REF _Ref479145595 \h (6) and  REF _Ref479252741 \h (7) that each of Efi is determined only by the ith row of A, i.e.  EMBED Equation.3 . Therefore, the effect of the predictor sparse structure can be examined for each row independently. For simplicity, in the discussion that follows, the subscript i denoting the analyzed vector component (the row of the predictor) will be intentionally omitted, but always keeping in mind that the expressions are given only for that chosen component (row). Two different approaches to sparse predictor determination will be explained. In the first one, sparse predictor is obtained directly from the optimal full predictor, by setting any chosen number of its elements to zero. In the second approach, the nonzero elements of the sparse predictor are reoptimized, resulting with an optimal sparse predictor for any structure. Partially zeroed full predictor Let us first analyze how prediction residual energy changes if a certain number of optimal predictor elements in the chosen row  EMBED Equation.3 are set to zero, while the remaining elements retain their original value. This operation can easily be described by subtracting the row sT from the original row  EMBED Equation.3 , where sT contains the elements that should be zeroed, as in:  EMBED Equation.3 .( SEQ ( \* ARABIC 8)S is a set of indices defining the elements that are set to zero in that particular row. The component residual energy based on row  EMBED Equation.3 , denoted with Ez, can be determined by substituting  EMBED Equation.3  for  EMBED Equation.3  in  REF _Ref479145595 \h  \* MERGEFORMAT (6), since this expression is valid even for a non-optimal predictor row like  EMBED Equation.3 . By utilizing the symmetry property of the covariance matrix C, it can be easily shown that the resulting Ez is equal to:  EMBED Equation.3 ( SEQ ( \* ARABIC 9)It is obvious that the residual energy is increased by (Ez compared to the case of the full row predictor as a consequence of zeroing elements defined by S. It is reasonable to expect that zeroing more and more predictor elements results in monotonic increase of the residual energy. To verify this assumption, differential increase of the residual energy, D, in two consecutive iterations is expressed, where in each iteration only one new element is set to zero. Therefore, row vectors sr-1T and srT defining the zeroed elements in two consecutive iterations differ in only one element, ad. It can be written:  EMBED Equation.3 ( SEQ ( \* ARABIC 10)For example, if s2T=[0,0,(3,0,...,0,(9,0], zeroing a new element ad=a5 in the third iteration results with s3T=[0,0,(3,0,(5,0,...,0,(9,0]. If (Ez is calculated for both rows sr-1T and srT according to  REF _Ref479148754 \h (9) with srT given by  REF _Ref479151316 \h (10), the differential increase of the residual energy is obtained as:  EMBED Equation.3 ( SEQ ( \* ARABIC 11)where  EMBED Equation.3 is row d of the covariance matrix C. For the example given above: D=(Ez(s3)((Ez(s2)=2a5(c5,3(a3+c5,9(a9)+c5,5(a52. In the first iteration (r=1), sr-1T is a null vector, so differential increase of the residual energy compared to the full row predictor is equal to cd,dad2 where cd,d is a positive diagonal element of covariance matrix C. Therefore, D is always positive in the first iteration. However, for any other iteration r, difference D depends on the row element set to zero in that iteration, but also on all elements set to zero in previous iterations. In that case, D is not necessarily positive since the first part of  REF _Ref479155808 \h  \* MERGEFORMAT (11),  EMBED Equation.3  can be both positive or negative. This means that setting an additional element of the row predictor to zero can even result in decrease of prediction residual energy although predictor has less nonzero elements. The above mathematical formulations were necessary to define the criterion for choosing the predictor element to be zeroed. This criterion is an optimal one if nonzero elements of the sparse predictor are exactly the same as those of the full predictor. The whole design procedure can be outlined as follows: determine the full predictor as in  REF _Ref478283845 \h (4) and initialize the binary predictor structure matrix to all ones; based on the current structure, form the matrix of differential increase of the residual energy caused by setting any nonzero predictor element to zero, by applying expression  REF _Ref479155808 \h (11) to all rows, and to all nonzero candidates ad in each row; from the matrix determined in step 2., choose the element (defined by its row and column) that causes minimal differential increase; set the predictor and the structure matrix to zero for that element; repeat steps 2. to 4. until the desired number of nonzero elements is reached. The explained design procedure belongs to the group of so called greedy algorithms. The criterion in step 2. is applied without any anticipation of the future, i.e. the element is chosen that is best suited for that iteration, without considering its effect on the iterations that follow. Therefore, the resulting sparse structures are only near-optimal. In our previous work two simpler criteria were also used in step 2. instead of expression  REF _Ref479155808 \h (11) and all three will be compared in this paper. The first one was based only on the right-hand part of  REF _Ref479155808 \h (11), cd,dad2, i.e. the influence of all components zeroed in previous iterations was ignored. For the second one, the predictor element with minimal squared magnitude, ad2, is zeroed in each iteration. Optimal sparse predictor Sparse predictors obtained by partially zeroing the full predictor are no longer optimal, i.e. they do not minimize the energy of the prediction residual. In order to become optimal, sparse predictors (their nonzero elements) should be recalculated. It is obvious from (4) that each row  EMBED Equation.3 , EMBED Equation.3  of the optimal full predictor can be calculated independently according to  EMBED Equation.3 , where EMBED Equation.3  is the ith column of EMBED Equation.3 . However, if any row of the predictor has some zero elements, then the optimal solution can be found from the modified set of linear equations, reduced in dimension. The procedure will be illustrated on the example of an optimal sparse row predictor  EMBED Equation.3 [ EMBED Equation.3 ] with only two nonzero elements. These elements  EMBED Equation.3 [ EMBED Equation.3 ] can be found as a solution of  EMBED Equation.3 as shown in Fig. 1. Generally,  EMBED Equation.3  is formed from C by removing rows and columns that correspond to zero elements of  EMBED Equation.3 . Analogously,  EMBED Equation.3  is formed from  EMBED Equation.3  by removing the same rows.  EMBED Word.Picture.8  Fig. 1. Calculation of nonzero elements of an optimal sparse row predictor Optimal sparse row predictors as well as differential increase of the component residual energy due to sparse structure can be determined using an efficient iterative algorithm. In each iteration of this algorithm, one predictor element is set to zero, while the others are recalculated. The sparse row predictor in iteration r, denoted as  EMBED Equation.3 is a kdimensional row vector with r zero elements in columns defined by the set Sr. Vector  EMBED Equation.3 can be found as in  REF _Ref479353053 \h (12) by left multiplication of  EMBED Equation.3 with an auxiliary k-by-k matrix  EMBED Equation.3 ={ EMBED Equation.3 }. Let  EMBED Equation.3 denote the jth column of  EMBED Equation.3 , while d denotes the element that is zeroed in the iteration r, i.e. Sr={Sr1,d}.  EMBED Equation.3 can be determined from  EMBED Equation.3 and d using the following recursion:  EMBED Equation.3 ( SEQ ( \* ARABIC 12)The algorithm starts from matrix  EMBED Equation.3 , that is equal to the inverse of the covariance matrix C. The resulting optimal row predictor for iteration r=0,  EMBED Equation.3 is equal to the optimal full row predictor  EMBED Equation.3 . Since all sparse row predictors found as in  REF _Ref479353053 \h (12) are optimal, the differential increase of the component residual energy between two successive iterations r-1 and r, as a consequence of setting the element d of the row  EMBED Equation.3 to zero, can be found according to  REF _Ref479252741 \h (7):  EMBED Equation.3 ( SEQ ( \* ARABIC 13)where  EMBED Equation.3 is the value of the zeroed element. Expression  REF _Ref479391681 \h (13) can be used as a design criteria for obtaining optimal sparse predictors. The design procedure is very similar to the one described for the partially zeroed full predictor with two differences. First, the criteria used in step 2. is replaced with  REF _Ref479391681 \h (13) and second, after zeroing the selected element in step 4., the nonzero elements of that row are recalculated according to  REF _Ref479353053 \h (12). Two simplified criteria based on minimal  EMBED Equation.3 and minimal  EMBED Equation.3  were also compared to  REF _Ref479391681 \h (13). SIMULATION RESULTS All of the proposed sparse predictor design techniques were evaluated and compared on the open-loop vector prediction of 10-dimensional Line Spectral Frequencies (LSF) vectors (Itakura, 1975) used in speech coding. The obtained results of the prediction gain vs. the percentage of zeroed elements are shown in Fig. 2. in two groups: for partially zeroed full predictors and for optimal sparse predictors. As expected, the optimal criterion for each of the groups results with maximum Gp compared to the simplified criteria. It is obvious that Gp graphs of the zeroed full predictors do not fall monotonically with the increase of the number zeroed elements, as opposed to the case for the optimal sparse predictors. Furthermore, the benefit of nonzero element reoptimization is evident. Finally, prediction gain of a full predictor VLP, Gp,full = 4.9854, is only slightly higher then for the optimal sparse predictor.  EMBED Word.Picture.8  Fig. 2. Prediction gain of sparse predictors for different design criteria CONCLUSUION The design procedure for sparse vector linear predictors that are near-optimal in structure and optimal in values of their nonzero elements is proposed in this paper. The increase of the residual prediction energy due to zeroing is mathematically formulated thus offering the exact criterion for structure reduction. The optimal criteria is also given for partially zeroed full predictor with non-optimal nonzero elements retained from the full predictor. Two design procedures based on the optimal criteria are given and compared on the LSF vector process together with two other simpler, empirically based criteria. REFERENCES Chen J.H., Gersho A. (1987), "Covariance and autocorrelation methods for vector linear prediction", Proc. ICASSP, 1987, pp. 1545-1548. Itakura F. (1975), "Line spectrum representation of linear predictive coefficients of speech signals", J. Acoust. Soc. Am., Vol. 57, Suppl. No.1, pp. S35. Petrinovi D., Petrinovi D. (1999), "Sparse vector linear predictor matrices with multidiagonal structure", Proc. of the EUROSPEECH  99, Budapest, 1999, Vol. 3. , pp. 1483-1486. Yong M., Davidson G., Gersho A. (1988), "Encoding of LPC spectral parameters using switched-adaptive interframe vector prediction", Proc. ICASSP, 1988, Vol.1, pp. 402-405.  This work was supported by the Ministry of Science and Technology of Croatia under project No. 036024.  Throughout the paper, vectors without transposition are assumed to be column vectors. Z\pN O X 3 4 5 6 7 : ; @ A W X Y Z [ \ ®® jzEHUj< CJUV OJQJmH mH jEHUj< CJUV jU j0JU 6CJH*6;5CJ55CJ5CJOJQJmH j0JOJQJUmH  OJQJmH ;XZ\N O $$09!7#$$$<XZ\N O 45ĿzupkfaZ  hi                 ! ;</01245<=?@ABDJK^_`acdefhկկկկ j EHUj< CJUVmH6H* j EHUjGZ< CJUV jEHUj< CJUV6 6CJH*5 OJQJmH mH jEHUjZ< CJUV jU=45,-"#$$$$09!7#$hij'('()*,-34GHIJZ[]lm jEHUjc< CJUV OJQJmH  j0EHUj < CJUVH* 6CJH*6mHjU jU56CJCJ5CJNHA,-"#$$$$4%&''''.*^***,,,,M-¼xqmhaZVQ    CD  ]  u<       7  P         |}  /02^_rstu}~,-./0129:MžjU5 jEHUj< CJUV jEHUj< CJUVmH j0EHUj< CJUVmHH* 6CJH*CJH*6 OJQJmH mH jU jEHUjt< CJUV5MNOPfgz{|}789:>?Aabdüj$Ujq$UCJH* 6CJH*6 OJQJmH  j!EHUj.< CJUVmHjZ!U jJEHUjx< CJUV jU j3EHUjx< CJUV<       "#########8$9$:$Q$R$e$f$g$h$p$q$r$$$$$$$$$$$$$$$$`%a%t% OJQJmH mH j+EHUj< CJUV j6)EHUj< CJUV H*OJQJB* j['EHUj< CJUVmH  jk%EHUj< CJUV jU5H* 6CJH*68$$$$4%&''''.*^***,,,,M-N.K234  & F $d($$09!7#$t%u%v%w%%%%%%%%%%%%%%%%%%%% & &&&[&\&o&p&q&r&&&&&&&&''''''''''''U'V'W'X'ƿװ jD OJQJmH mH jZ7EHUj:< CJUV5 jh5EHUj4U j2EHUj%< CJUV j1EHUj< CJUV 6CJH*6 jU j/EHUj< CJUV7X'Y'''(()) ) ) )))))***.*0*V*X*Z*\*`*b*****************(+*+,+2+4+6++++++++++++++++++,,,,,,,,, jD ja6 OJQJmH mH jj:EHUj>< CJUV jU 6OJQJ H*OJQJCJH* 6CJH*56CJI,*,+,A,B,C,D,E,F,G,M,N,O,P,Z,[,q,r,s,t,v,w,x,,,,,,,,,,,,,,,,,- - - -----.-N-O-R-S-T-U-V-W-X-Z-[-\-]-^-_-`- j-CJH* jD6 j|BEHUj < CJUVmH OJQJmH  j>EHUj< CJUVmHj>U H*OJQJ 6CJH*5mHj=U jU>`-a-b-.... ...... .(.*.,...6.8.@.B.D.F.H.~.......|/~////////////0000b0c0p0q000-1.1D1E1U1V1Z1[1]1^1q1r1s1t1333 jDEHUj< CJUVjjDU jUCJ H*OJQJ 6CJH*CJH* jCJ6 6OJQJCJH*5GM-N.K234 666&7;;@@ ADDDDE/GGG`GaGJ+JMMM,N8NPP½yupmhc^YmSm  6STab  {      ! N      "    ?3333333f5h55555555555888888999i9j999999999999::::>;@;B;D;;;<<======&='=(=)= jJKEHUjx< CJUV j.IEHUjg‘< CJUVmH CJH*CJ6jHUj4HU 6CJH* 6OJQJjGUmH jUj:GU>4 666&7;;@@ ADDDDE/GGG`GaGJ+JMM$$$09!7#($<  & F )=u=v====================>>>>>>>>>>>?1?2?E?F?G?H?I?J?]?^?_?`????} jYEHUjVܑ< CJUV jWEHUjIܑ< CJUV j:UEHUjܑ< CJUV j4SEHUjۑ< CJUV jaQEHUj}!< CJUVmHH*6 j\OEHUj‘< CJUV jZMEHUj‘< CJUV jU0???????????? @!@4@5@6@7@F@G@Z@[@\@]@m@n@@@@@@@@@@@OBPB]B^BqBrBsBtByBzBBBBB jfmEHUj< CJUV6 jleUj V< CJUV jcEHUj< CJUV jaEHUj< CJUV j_EHUj< CJUV5 j^EHUj< CJUV jU j[EHUj~ܑ< CJUV1BBBBBBBBBBCCCCCCC3C4CGCHCICJC\C]CaCbCjCkC~CCCCCCCCCCCCCCCCCCCCCCCCCCƑ jyEHUH* jwEHUj< CJUV juEHUjR< CJUV jsEHUj< CJUV6 jqEHUj< CJUVmHj{qU j~oEHUj4W< CJUV jU 6CJH*6CC'D(D/D0D2D5D6D8D9D;DMM,N8NPP4QQhS`TT!U"U#U$U%U&U'U(U)U*U+U,U-U & Fx<P4QQhS`TT!U"U#U$U%U&U'U(U)U*U+U,U-U    n    QQQRRRS8TDT`TaTTT-U j0JU6 &P . A!"#$%zDdt B  S A? 2^An\L8RDn`!^An\L8R q8xSJ@?I"O衅zS(B=ؾ@` -h=7zG/wqgmDk̷7]@{C5'&AlbA}_1/wJ!;EӸ4Aέk3u9,=a e"M*䤡R4j+G|ݧ(UX i'ε}rM+)R|w=f\n$~h²m[1 fl+6o)ܠygǴ{h_{WZz.A%v 2ҸW쩩8yQC|9z8g+kD$Zծ{ֺdjAƸ&R4~I= 虖ItGFc ]}Dn0$ =Dd@B  S A? 2ClAkn`!{ClAk@ h IxJ@ƿ4 O[лɛElOzL+E(>B/^|o<3R I6i`f̷B p.`<^%6^L$8%R512O >Wn3uoz76c}j|WNo~F0"4# .DdB  S A? 2p-:|w :XLn`!D-:|w :Xd)xTK+Ag$&&TCA%`#V*h 'C 4沈A$ b (" bc>= 73;;ˠ:F f$dL0 ժ`XJ0d:/`1n4T q`AOՅ\i(\="c/m⸎"nWs +NZMinJsٸEkLq""8i}TT >ފZ6iщR~(kc߬ C6+)v>` jsywuלb~W@FB/}9~LYrG+^.ok'}0asmV'~;fVdC] Gx{aSbb'Bf'g|* mY*kņE/[2Y5Ǧ)g0$3ξ;ѕ!O5}_ ϲo9cASd:NyCmbF69QZYu%tw DdB  S A? 2"׾bh6dqdn`!\"׾bh6dq^ *xQAJP}3i " WZѽ`t",6VlQʵpap!x[+& 7yx3Ch&ES7dj&H+EQhCZbXA!M ~4YnbBoYN4.H]Ïzi2 xDls큟B^/⠢ NYU$]]f)SJ%XX~*Syn ࿜ՄS+bPTo_D{P#x6Oۭ{)1v(I1ztVDd B  S A? 2 %qƺKc 1_ n`!W %qƺKc 1 028%xOA߼ݖ *Eƃф0 p&%H&ȁ+$z d/^C3ڡoHf|{y;N@}ԝD=rՍBAQq̌h؊ |ݎ:5RҤ k %o|b'a\ b=.,UUU$^G O]-ak#CXYL$qӐ4q ]\}V@#1P}]DdB  S A? 2)9jB4> Z 8 n`!9jB4> ZRxcdd``e```beV dX@`FE%"LPL0}?@ABCDEFGHIJKLMNOPQRSTUVWXZ[\]^_`cfghjiklomnpqrsutvwxzy{|}~Root EntryE FR=@SjeDataEtPGE Y7WordDocumentdExEN"ObjectPool55E,@hE,_1016002829&FDjDjOle ?CompObj%'@fObjInfo(B FMicrosoft Equation 3.0 DS Equation Equation.39qޘ8II e i (n)=x i (n)" iArial e~T  x(n"1) . FMicrosoft Equation 3.0 DS EqEquation Native C_1016259939[+FDjDjOle FCompObj*,Gfuation Equation.39q,x~IuI  iArial e~T FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfo-IEquation Native JH_10161000180F\Lj\LjOle LCompObj/1MfObjInfo2OEquation Native P_10162746965F%]j%]jbIfI E fi =e i2 (n) n=1P " =E 0i "2 iArial e~T   i + i e~T  C  i   ,     E 0i =x i2 (n)  ,     n=1P "  i=1,& ,k.Ole XCompObj46YfObjInfo7[Equation Native \8 FMicrosoft Equation 3.0 DS Equation Equation.39q~IzI  i FMicrosoft Equation 3.0 DS Equation Equation.39q_1016259977:F˂j˂jOle ]CompObj9;^fObjInfo<`,x~IuI  iArial e~T FMicrosoft Equation 3.0 DS Equation Equation.39q8bIfI C i = Equation Native aH_1016101089.?FjjOle cCompObj>@dfObjInfoAfEquation Native gT_1016101110~DFƃjƃjOle i i FMicrosoft Equation 3.0 DS Equation Equation.39q(rIUI i=1,& ,kCompObjCEjfObjInfoFlEquation Native mD_1016002862IF9j9jOle oCompObjHJpfObjInfoKrEquation Native s FMicrosoft Equation 3.0 DS Equation Equation.39q޴ITI E fi =E 0i " iArial e~T   i    ,   i=1,& ,k._10162600238QNF҂j҂jOle wCompObjMOxfObjInfoPz FMicrosoft Equation 3.0 DS Equation Equation.39q(I(I  Arial e~T FMicrosoft Equation 3.0 DS Equation Equation.39qEquation Native {D_1016260056SFlj`jOle }CompObjRT~fObjInfoUEquation Native D_1016108815oXF@]j@]jOle (I(I  Arial e~T FMicrosoft Equation 3.0 DS Equation Equation.39qӨaIfI  z  ACompObjWYfObjInfoZEquation Native _1016260100L]F`j`jrial e~T =  e~T "s  e~T        ,         e~T =  j []   ,  j=1,& ,k s  e~T  = s j []   ,  j=1,& ,k            ,         s j = j   ,  for j"S   0 ,  for j "S{oma FMicrosoft Equation 3.0 DS Equation Equation.39q,LrIuI  zArial eOle CompObj\^fObjInfo_Equation Native H~T FMicrosoft Equation 3.0 DS Equation Equation.39q,LrIuI  zArial e~T_1016260117bF@j@jOle CompObjacfObjInfodEquation Native H_1016260133`rgF/j/jOle CompObjfhf FMicrosoft Equation 3.0 DS Equation Equation.39q,lII  iArial e~T FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfoiEquation Native H_1016019702lFXjXjOle CompObjkmfObjInfonEquation Native _1016112702qFij`rjt,vII E z   =   E 0 "2  Arial e~T +   e~T C + s  e~T C s    =   E f +s  e~T C s   =   E f +E z   .Ole CompObjprfObjInfosEquation Native l FMicrosoft Equation 3.0 DS Equation Equation.39qPxaI@fI s rArial e~T   =   s r"1 e~T +s  e~T   =   s r"1 e~T + 0 , & , 0 ,  d , 0 , & , 0 []  . FMicrosoft Equation 3.0 DS Equation Equation.39qބ,vII D  =  E z (s r )" E_10162748253vF|zj|zjOle CompObjuwfObjInfoxEquation Native _1016275148{F`Ej`EjOle CompObjz|f z (s r"1 )  =  s rArial e~T C s r " s r"1 e~T C s r"1   =  2 s  e~T C s r"1  +  s  e~T C s   =  2" d   d e~T  s r"1 +c d,d " d2     , FMicrosoft Equation 3.0 DS Equation Equation.39q,,vII  dArial e~TObjInfo}Equation Native H_1016116761VF6j6jOle CompObjfObjInfoEquation Native _1016185447Fjj FMicrosoft Equation 3.0 DS Equation Equation.39qhȟI pI 2" d   dArial e~T  s r"1 FMicrosoft Equation 3.0 DS EqOle CompObjfObjInfoEquation Native luation Equation.39qPIxI  Arial e~T =   i e~T FMicrosoft Equation 3.0 DS Equation Equation.39q_1016185475F׃j׃jOle CompObjfObjInfoEquation Native <_1016185538FjjOle CompObjf DII C=  FMicrosoft Equation 3.0 DS Equation Equation.39q(ITI  = iIObjInfoEquation Native D_1016275325y|FjjOle CompObjfObjInfoEquation Native 0_1016191881B)F`j`j FMicrosoft Equation 3.0 DS Equation Equation.39qd~II   FMicrosoft Equation 3.0 DS Equation Equation.39qOle CompObjfObjInfoEquation Native P4I}I  sArial e~T = FMicrosoft Equation 3.0 DS Equation Equation.39q`|I4I " 1 , 0, & , 0, " k_1016192005F@U"j@U"jOle CompObjfObjInfoEquation Native |_1016192073F 3jE?@ABCDEFGHIJLPUXYZ[\]^adefghijmpqrstwz{|}~uation Equation.39q<I8I " 1 , " k FMicrosoft Equation 3.0 DS Equation Equation.39qOle CompObjfObjInfoEquation Native X<J4I "C " s = " FMicrosoft Equation 3.0 DS Equation Equation.39qI}I "C_1016193691FAvjAvjOle  CompObj fObjInfo Equation Native 0_10161937191F j jOle CompObjf FMicrosoft Equation 3.0 DS Equation Equation.39q,IHI  sArial e~T FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfoEquation Native H_1016193780F`ӗj`ӗjOle CompObjfObjInfoEquation Native 0_1016193794F`j`jmI}I " FMicrosoft Equation 3.0 DS Equation Equation.39qIzI  [$@$NormalmH Ole CompObjfObjInfoEquation Native ,_1016223245c6 F`=jjData 1Table CompObjKhDd0B  S A? 2!*:1JRL]Dv`!!*:1JRL]  e%Hx?hQ{w]l-* )\@%P%Ɗ h BQ[A]t" RZ] &{wwwh`q1-Θja|4lVĜe> jP#4 8{"3/1٘fz9#|u[- D'~EČ;O=pѡNltd@{yio Uf~E:ёf~5^Sf=o+E:ё<y>0聋tb# y>||~oNc>:FG |h8ME:ё<|ԑlE5t+s훵 5_XOty(wޣ]Jϊ2QvƳyg߲"gdR%3Cpʞ-[f68eelk4kz7Z3{<[ދ2S/}s"\M\7k?w{IuzǦ߭`qk T9frq"Wcw8I|FT,_%rM ^KwR*hDdTB  S A? 28P N/7v`!8P N/7  ȽXJtxcdd``a!0 ĜL  312Ec21BUs30)0)0Qcgb  P#7T obIFHeA*CT f0 PeDd @B  S A? 2f̉D~lBB/v`!:̉D~lB8 xJBQWK HK wJoպTUA!& zߣ=:͌G qӁ 3sUKX@D!vYI5\12ouLL)THdK ", zN6I9Xˏr# d'O80U' @|gLjDPҭm'LH,폇ୣv́ɥ# "{MgIաZODdt B  S A? 2FqM C‰ex}"+v`!qM C‰ex} q`\xJ[A93&& ucB,%vZ)B6vQH6B] Rk7nvcUAsL&^k{2o B @`2-EӦAzvd'HЩG.CΏ=0LAAp`zM;q TnMX٥ys띷@?oFbzeɵe-u|W^?ap4s~%-kLJZ? ~t"'<=OtpΠMzu/SlHhwOE5,#M¼ wt#s?a;*7\q|OǓ{&Қ@s%Zcx8$WmkA-cr~]QUMmW8ti{q$-9K$zkVpJY pe1DdB  S A? 2j$25?On!w v`!oj$25?On!h=xR1KPM FA .Vtw:ظ7B14͟`@gq/k1 &{߽ Xg0֠:VDR!.B*-Z>a7x!?VU  ¹2]ET4~8l0N4+dv ,ӓݼiZ*EE[JGXھw|;2& d^0o;3 3 t>  tt @LP@GTimes New Roman5Symbol3& Arial"hDD!0WDavorka PetrinovicDavorka  FMicrosoft Word Picture MSWordDocWord.Picture.89q FMicrosoft Equation 3.0 DS EqObjInfo.MObjectPool ejj_1015155910F ej ejOle NCompObjOfObjInfoQEquation Native R$_1015155921F ej ejuation Equation.39qkIvI  FMicrosoft Equation 3.0 DS Equation Equation.39qII c 1Ole SCompObjTfObjInfoVEquation Native W1 c 12 ""c 1k []c 11 c 12 "c 1k c 21 "c 2k """c k1 c k2 "c kk []"[]_1015157405F ej ejOle _CompObj`fObjInfob FMicrosoft Equation 3.0 DS Equation Equation.39q,vI|I c 11 c 12 ""c 1k c 21 ""c k1 c k2 ""c kk []"a Equation Native c_1015157812F ej ejOle kCompObjlf1j 0"0a kj []= 1j 0"0 kj [] FMicrosoft Equation 3.0 DS Equation Equation.39q(kIvI c 1ObjInfonEquation Native oD_1015158018F ej@jOle u1 c 1k c k1 c kk []"a 1j a kj [] T = 1j  kj [] FMicrosoft Equation 3.0 DS Equation Equation.39qCompObjvfObjInfoxEquation Native y_1015158965F@j@j,vI|I c 11 c 12 ""c 1k c 21 """c k1 c k2 ""c kk []"a 1j 0"0a kj [] T = 1j  2j "" kj [] FMicrosoft Equation 3.0 DS Equation Equation.39qLrIaJ c 11 c 12 ""c 1kOle CompObjfObjInfoEquation Native  c 21 """c k1 c k2 ""c kk []"a 1j 0"0a kj [] T =g 1j g 2j ""g kj []_1015159056F@j@jOle CompObjfObjInfo FMicrosoft Equation 3.0 DS Equation Equation.39q(PSJ J c 11 c 1k c k1 c kk []"a 1j a kj [] T =g 1j g kj []Equation Native D_1015159136F@j@jOle CompObjf FMicrosoft Equation 3.0 DS Equation Equation.39q<leJI C""a j = jObjInfoEquation Native X_1015159421F@j фjOle CompObjfObjInfoEquation Native _1015164248F фj фj FMicrosoft Equation 3.0 DS Equation Equation.39qܘPSJI "C""a jT =" j     !    "a j       FMicrosoft Equation 3.0 DS EqOle CompObjfObjInfoEquation Native uation Equation.39q,vI|I "C"a 1j a kj []=" j     !     "a j       FMicrosoft Equation 3.0 DS Eq_1015164674F фj фjOle CompObjfObjInfouation Equation.39qkIvI "C FMicrosoft Equation 3.0 DS Equation Equation.39q XIpI " jEquation Native 0_1015164737F фj фjOle CompObjfObjInfoEquation Native <_1015164954F фjڄjOle  FMicrosoft Equation 3.0 DS Equation Equation.39q<t~IzI C""a j = j FMicrosoft Equation 3.0 DS EqCompObj fObjInfo Equation Native X_1016184954 FڄjڄjOle CompObj fObjInfoEquation Native uation Equation.39q@yII c 11 c 12 ""c 1k c 21 """c k1 c k2 ""c kk []"0"0a kj []=g 1j g 2j ""g kj [] FMicrosoft Equation 3.0 DS Equation Equation.39qII c 11 c 12 ""c 1k_1016192249FڄjڄjOle CompObjfObjInfoEquation Native _1016192477 FڄjڄjOle CompObjf c 21 """c k1 c k2 ""c kk []" 1  2 "" k []= 1  2 "" k [] FMicrosoft Equation 3.0 DS Equation Equation.39q 4IeJ C"= FMicrosoft Equation 3.0 DS Equation Equation.39q xII C"=ObjInfoEquation Native <_1016192542FڄjڄjOle CompObjfObjInfoEquation Native <_1016192674$!FڄjڄjOle CompObj "fObjInfo#Equation Native 0 FMicrosoft Equation 3.0 DS Equation Equation.39qInI c 11 c 1k c k1 c kk []"" 1 " k []= 1  2 [] FMicrosoft Equation 3.0 DS Equation Equation.39q8tI J "C " s ="_1016192740)&FjjOle CompObj%'fObjInfo(Equation Native T_1016193105+FjjOle CompObj*,f FMicrosoft Equation 3.0 DS Equation Equation.39q xII C = Oh+'0`   ( 4@HPXssDavorka PetrinovicavoObjInfo-Equation Native <WordDocument/SummaryInformation(0     MO !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLNPRQSTUWVYX[Z\]^`abcdefijklmnopqrstuvwxyz{|}~Y bjbjWW  == ]***ZZZZZ f Z~~~~$ \*!l~E~l~^*>L,*r MFꜿZZ  EMBED Equation.3  EMBED Equation.3   EMBED Equation.3   EMBED Equation.3   EMBED Equation.3    &'(*=>?@STUhijk꓉~j CJEHUmHjޑ< UVmHmHjCJEHUmHjޑ< UVmHjCJEHUmHjQ< UVmHjCJEHUmHj < UVmHjCJEHUmHjܑ< UVmHjCJUmHCJmH jUmH# KQST   KQSThijN N!,"#!$6$% Normal.dotrDavorka2voMicrosoft Word 8.0@@*#.ꜿ@*#.ꜿ ՜.+,D՜.+,4 hp   FER-ZESOI-j  TitleDocumentSummaryInformation8_1016203158@3FjjOle CompObj24f "%(+,-./0149<?BEHKNOPQRSTUVY\_bcfikorux{~ 6> _PID_GUIDAN{99A6534E-FF2A-11D3-9D72-0040054232E7} FMicrosoft Equation 3.0 DS Equation Equation.39qP|IqI ( sArialObjInfo5Equation Native l_10162235408FjjOle   e~(r~) )  e~T FMicrosoft Equation 3.0 DS Equation Equation.39q4II  sArial e~(r~)CompObj79 fObjInfo: Equation Native  P_1016203483T=F`6j`6jOle CompObj<>fObjInfo?Equation Native @ FMicrosoft Equation 3.0 DS Equation Equation.39q$I}I B (r) FMicrosoft Equation 3.0 DS Equation Equation.39q_1016203346BFQ jQ jOle CompObjACfObjInfoDEquation Native L_1016203426^GF%j%jOle CompObjFHf0xII b i,j(r) FMicrosoft Equation 3.0 DS Equation Equation.39q(XI$I  j(r)ObjInfoIEquation Native D_1016204311;OLF ?j ?jOle   FMicrosoft Equation 3.0 DS Equation Equation.39q,xII B (r"1) FMicrosoft Equation 3.0 DS EqCompObjKM!fObjInfoN#Equation Native $H_1016204327QF Xj XjOle &CompObjPR'fObjInfoS)Equation Native *uation Equation.39q(ITI B (r) =B (r"1) "1b d,d(r"1) " d(r"1) ( d(r"1) ) Arial e~T   ,    s(r) =B (r)    ,   for  r=1, 2, & , k  . FMicrosoft Equation 3.0 DS Equation Equation.39q$I}I B (0)Й`Hb@Cx=K@ǟ{RmDQ *P7?"64BBӊ-h7'q\]tMK+5y!@0 J#9KApWʄQIuyQ՚tf E` YW [+('n{xUVN',.UXNZwd4}p+E<++SA&_3ۊK.+0%/*~4^i=.qk۪tf(􁼳o\Ԣ٬ ;#w,|NyYP|>)X UEJzV\)VJY|G534SwXJ99\{֣49 G=-#|j{>!Dd@B  S A? 2ZD!yL56n`!.D!yL5 @CxmAJPE{&PDYtvPf0Ш؀d\[>d c4{}~0FߩĪDcTХ*Lpr n{縂!rod4J8Io)PaH[0{,];QtCk-7]\gyfG4͓j ?icdHv.- <%w /DO_S*vvL@Dd B   S A ? 2UN <(gdn`!UN <(gdp-F8xMLAn[]*UPQ* M('ㅚSj7rBc5$됗ʕKA*Oϕ[Owt+&H}d>uzUtK6_)¤z5UOΫ X3b&&S|f%4?CϘ..fmwW~4K~ԥf~),T]{>4_5A]Ѹ1W]Q75FٙϨAB̈́g[D&DfIgAmT:N%,OlFcifP;]+Z9XwAv_bDdhB  S A?  2A,BOktn`!,BOkd@H|xm AAƿsK)YŽ <[,.#x`mqE)K+יc1u3|72@QY .nT"U(RIeUZL֝\;RQz2 "3WΡt@RYɡ;^M4ç q;޲πIҐmLN-!IG[_R@M/Hy]~1qz3 1k;jDdB   S A?  2)9jB4> Z Kn`!9jB4> ZRxcdd``e```beV dX@`FE%"LPL0}d c4{}~0FߩĪDcTХ*Lpr n{縂!rod4J8Io)PaH[0{,];QtCk-7]\gyfG4͓j ?icdHv.- <%w /DO_S*vvL@}DyK _Ref478283845DdhB   S A ?  2!Ldz]wn`!U!Ldz@|#xcdd``dd``beV dX,XĐ Ɂi A?d-@P5< %! `f: XB2sSRsRA.Jsi#Y@j3|m&bb= .fB`?~h616 b#$#IĻ3.7yC!IIwuf&pI9 L\oO.p5S`|Ĥ\Y\M`u+jDd,B   S A ?  2z T^hˉ `LVn`!N T^hˉ `L ~ xcdd``fd``beV dX,XĐ  A?d@eKjx|K2B* R. XB2sSRs@.Jsi#FVl8_0`?~@2y& H 2b7D(ȃb#fbop8I90.-4L*LP,PLcdbR ,.Iex&f:nL}DyK _Ref479145595Dd B  S A ? 2|-^_Or"n`!|-^_Or`` xJAgy AV+#>El$)OsFs\-AsYFӞ3/+Ljo=USɛ،x ߵZmH{E|i2ݛ3 _S ӠGM~섈}DyK _Ref479145595}DyK _Ref479252741Dd@B  S A? 2ZD!yL56%n`!.D!yL5 @CxmAJPE{&PDYtvPf0Ш؀d\[>d c4{}~0FߩĪDcTХ*Lpr n{縂!rod4J8Io)PaH[0{,];QtCk-7]\gyfG4͓j ?icdHv.- <%w /DO_S*vvL@Dd@|B  S A? 2Ef|ajY uSr!'n`!f|ajY uSrZ` 0xEAJPTۤP Y7, 4@# U@psE o>xw!zV%Rʂ19܈_|%3B(IJDK23cR[ ,0@J޳dnEtM+y6ey*ɅqEmlPҦ& ]U{M}Q9ꎪ$x4<Dd@|B  S A? 2Ef|ajY uSr!z)n`!f|ajY uSrZ` 0xEAJPTۤP Y7, 4@# U@psE o>xw!zV%Rʂ19܈_|%3B(IJDK23cR[ ,0@J޳dnEtM+y6ey*ɅqEmlPҦ& ]U{M}Q9ꎪ$x4<DdpB  S A? 2otъyu xVKU+n`!Ctъyu xV 0HJxOARJ NAhb҅|#`/.mqZʪ ,B./wnFW rW=)7q'd'489?>tGD&:fPZ7&=Wb6٩A #=t8;rЪA9tJu4y3n`!/)GP(e @CxmAjPE{iD qZBP6 0@#g]BFYCw s_~*O޿{|~r|ӅUT!6ƨ2UMeo]acC^Tʩ$<=<'xH: EEڂY2wD)鎬%{չ0&G4}ϓ7`SHIҌӑf9Ri$r$m+*n#}K̷/b؞w6jA+}DyK _Ref479145595Dd@B  S A? 2\T=ɏD Ή685n`!0T=ɏD Ή6 xmQjPELZQhWP_\ PDPhlB t ]B~d _M^Zwpjh~rjl BlQG739w\?ܔr6 !{.S&'@ui`9<2i F8xYďѝW#꒵DQ EQNx l )Kq_3oH_v?R-$r$p m+*n?;j؟@DdB  S A? 2zvYPȄ2V7n`!NvYPȄ2(]?xKTQǿQʞJ9`JIEhDQq5Q6Ѵ ڷA-qADwޕ󜰨~ί=2C_2īH)cQEL~c9G8/O> zu8 -jYh\3&ͥ^Fw4B@?\`nRZqOd4[ /FYTd( -cK؋:yۗ ksb#-9Ky%U*ۙJf@8,ԁSQιx{yͣgO>vNqs|_˰?y|rcmԺsP1N2_p1 3/aNvq/I= ATu8%c?'KA?5~s@ z״b2v4m\!Jō lJCuL?t<]\0c9 ]͚oٮG^;hxV,W3q;.,DdB  S A? 2l2CNyj$F=o#cr:n`!jl2CNyj$F=o#c^818xkQnk7v7?jbxPM5EJ=ILn6$[gK9=^'`cE-Kyov8y o_;_Ș\aDZ\Ya%@Dgqǧ8ݙ(C,Lr;ql3Aak gYcnsmlo]|b艐;SFUޕp&~Վž3|ގ-p\ĒLR;U)?SBIU=4(S]z;K4uKlzXE)P~_{4SQ֒;q 2j%ˌxT:ԧm'd[)OQ65G3U>v&<~.r[om@1&ῦ4}|rOsɡlS^P~F K9d^?ޒqx<*}DyK _Ref479148754}DyK _Ref479151316DdpB " S A? 2V x}FΌ;(鵟2>n`!* x}FΌ;(鵟* %S:xOAmi˶P+D5r5w+4CԨăH( zDhԽ LwԀ4Ͼy3%n);s.%_!T*+;I_餤n^Nۣ_ $Ēr蜽pg:&s,n*EWx!p*(X#dxRX.s$M+PЃL6H9(60EJfdVu4&}wJK-ےj{+%GQ));yòp ݂%dBJ:O{l ̴!%+{Wz#iAO7Qu}]ƜJ` xw~rMTvkP:`#@*hDd@B ( S A? 2Xt͛*Ii -4Bn`!,t͛*Ii -ˤ @CxmMJPϽ 3 4`J'%AB:-yKi|>}`~r|ӑUT!6ƨ2sU;>qL).`lj92.[>EԹe-wȞ]+QOZRmfiQUry@JړfGGzsY}To. P_<!#QFwc9i0J껢CP׬. $3}DLnr"|kh\bZYv)?EJwDi'.}KxEK+vSEx񴳈iifƧqa잝vkM*qAftJ=dFw 1oz`dI}}DyK _Ref478283845}DyK _Ref479155808}DyK _Ref479155808}DyK _Ref479155808DdB ) S A? 2X%fjoǛbrIn`!ZX%fjoǛN@C(xOJPƿjӂA + w,bsDh؂dgo . q%?=73|7 8{0|ӕ:L⪪3M0Yu.!:PO.%+$"Ţ['"̯ RR##uxbKwMC7ǎmCu+|o8Ts0͒Y0Jni_HJ~m?(9[Kz2E&%Czi!+crqO'࿜]/5b7as%1o<ɰQWFDd,B * S A? 2z T^hˉ `LVKn`!N T^hˉ `L ~ xcdd``fd``beV dX,XĐ  A?d@eKjx|K2B* R. XB2sSRs@.Jsi#FVl8_0`?~@2y& H 2b7D(ȃb#fbop8I90.-4L*LP,PLcdbR ,.Iex&f:nLDd@B + S A? 2lg{9y>SPoTHMn`!@g{9y>SPoT8(+ xcdd``ad``beV dX,XĐ i A?dǀjx|K2B* R.= s01d++&1,\ r. mĕL`fF> 2əx|b#$'FILHS$IgMa`p\2pAC 3`@``G#RpeqIj.;t71[DdhB , S A ? 2ouj`ػ2KOn`!Cuj`ػ2`@P|xcdd`` @bD"L1JE `xX,56~) M @ k+x UXRYvoaA $37X/\!(?71(%@i m 3Hm&Ԅ {@.[̈́P~7_ A4927)? "yp{EsX?^a.hqlu cdbR ,.Iex:f: S{Dd,B 6 S A(?  2=!og6>Qn`!!og6>dxm AaRnJfaG^d#`m+X[(9JwQ_*\upC` }o6vK,5K7š4Bi$v $wvs<׷`:DdB . S A!? !2p޷xu"LxSn`!D޷xu"`:@CxmPMJ`R +[н !`b W=\*^ۂ~" ||/|3/%:u%29gȐ 9g>72nNE俑@&q)Rc.ۂ,#Ut}o-åѡÈ7H?NH5_r~V(y/ݴy`D)3Ӕީm𮛩79͙CBжh|; b>U,/.0THPDdhB / S A"? "2 tH*ǎ~Un`! tH*ǎ @ |\xRNP =v%£11| *D"h(ha'023 2+J :iDR ]>>5}5TʩD!P<!:S̮<njҊvKD`)(]|J̩\JSf^HW1,hw|NUX<բJ)+ou(d$Hw/g}=sEĊ+᪠ɻ⫡iF-Vs?w;O&m[f25#^Ek7YT:eD(~i Az'ļt{I (g DdB 0 S A#? #26J)BڗfWn`!^6J)Bڗ&`:@C,xmQJ`ԟ"x"NJyDXh4G*K^0Ҫ$,3Bp `<rl&qYҡv.,S=6wpҐTB[F/9BE.[ `yAMau`s 0#I)$5536<DdB 9 S A+? )2) cn`!@xcdd``e```beV dX@`FE%"LPL0}0*DdB : S A,? *2d6c!;9Ї6@en`!86c!;9Ї6rPLxX]lTUwCiKEmE԰5>K_xhZK%HRFy0BXMPMRdECUƒQJ=gnﶻ]6{w̜9s̙l  YlU0)~\gֻ)=08y()qHǩ䐭=C\њP}xHtoO:DvERM uvR3#򟯄Aɓ|zyIԌQ5-#,CG )r0DЪpN`lxߡf:]RkeBfSLT@:Ceb0`^rW$gZt ݂[xeŜ:-B%51,#EU%%4Xi6>-#d]tКԩF5We2߬E-'ZGsa MںرSZvegW(PپsWۈڃ I',38xQuH3?aqak<̯e֙p 0 805vkݕ1NXg8)q$k]{GpaqakCc+0 805verG^QSܯڷ"ʳ ӒH|cs02"l/|}&F2ɳq/cgΒJOVc~G%*&Y1Ʊ w97Βs4Q#t'ź"wl3¸+yĖ?P]ŷ59NFg<\ntOmk(Ioj.C.lܧcｘWF%NiC~&΀zC&'#Fedѻ53ekD^&_ҙB&l2 7m!ڧHpٿX{ڙ^[v}wti9FߧؘѷP2*NiE, %h(r i߬ԝUyo-ʟY09ͤt9L')]fLJDž>h.ZY+]VLfcG u+R;e޸uCc]RHB-)ʵ(bWu D9ɗ/ʝOOpnnlm:6/{,HvldG}YH'Y)yLWET: xMWkJ ʮvE2hu|:ٯhZyi`{L=*F/C_}~/t^OV_$ߴYA9f/@ ;Qp'h9{[u ].g޺m5xܡخcsOw\cC o-ح5v)v؝Ocw>\֥y}}Z2ohڗEq^N'nVٱ$f/C o5>ExUFp5Wy,98ǽUp"Td~4xc6T1O}9` `klHY^>*6D͟{p5]Bt_ԊdOks+=orBLP ;}vQ77+1=՝HFz7+?2wv@G˅ b>,wBEMg}DyK _Ref479353053DdB H S A+? -2) 0*DdhB = S A/? .2O[B l%I kN+sn`!#[B l%I kN@|xcdd`` @bD"L1JE `x,56~) M @ k!} UXRYvoav@Hfnj_jBP~nbC#%@y mv&k`HDF? |jRԤ=L1j $'.߹$8~3 0#LĤ\Y\u`uCEIDdB > S A0? /2bCͶ7wLXt&qU>un`!6CͶ7wLXt&qU @Cxcdd`` @bD"L1JE `x0 Yjl R A@ nP5< %! .t! `bM-VK-WMc`hD(\фҀk|  $*z&[ ~&Ԧmb5IlR4 pf``b#RpeqIj.;t3M"JVDdB ? S A1? 02\]; AJP8wn`!0]; AJP @Cxcdd``~ @bD"L1JE `x0 Yjl R A@ nP5< %! .t! `bM-VK-WMc`Xr.PHq2ʀ] 1&20H&AL`~{@a n5Id#.hXp ipf``GI)$5kI DdhB @ S A2? 12O[B l%I kN+yn`!#[B l%I kN@|xcdd`` @bD"L1JE `x,56~) M @ k!} UXRYvoav@Hfnj_jBP~nbC#%@y mv&k`HDF? |jRԤ=L1j $'.߹$8~3 0#LĤ\Y\u`uCEIDdhB A S A3? 22O[B l%I kN+{n`!#[B l%I kN@|xcdd`` @bD"L1JE `x,56~) M @ k!} UXRYvoav@Hfnj_jBP~nbC#%@y mv&k`HDF? |jRԤ=L1j $'.߹$8~3 0#LĤ\Y\u`uCEIDdhB B S A4? 32k%Y!8,OMG}n`!?%Y!8,OM@@| xcdd``fgd``beV dX,XĐ ɁɁERcgb Bx UXRYvoav@Hfnj_jBP~nbC#%@y m8X |/F_4P 27)? b.#اVÅ]1n!"I P0+L@&91pAÒ @Sa;F&&\wc!MZDd$ B C S A5? 42F_Ђ%­o"_1016204057VF=aj=ajOle 2CompObjUW3fObjInfoX5Equation Native 6@_1016204581m[FhjhjOle 7CompObjZ\8f FMicrosoft Equation 3.0 DS Equation Equation.39qP$IPoI ( sArial e~(0~) )  e~T FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfo]:Equation Native ;l_1016204567J`FyjyjOle =CompObj_a>fObjInfob@Equation Native AD_1016206114YheFpjpj(II  Arial e~T FMicrosoft Equation 3.0 DS Equation Equation.39qXII ( sArialOle CCompObjdfDfObjInfogFEquation Native Gt e~(r"1~) )  e~T FMicrosoft Equation 3.0 DS Equation Equation.39q4@yII E s(r"1) =E 0 "( s(r"1) ) Arial e~T _1016208193jF`aj`ajOle ICompObjikJfObjInfolLEquation Native MP_1016206012oF`˼j`˼jOle WCompObjnpXfE s(r)  =E 0 "( s(r) )  e~T }   !   D=E s(r) " E s(r"1) =  1b d,d(r"1)   d(r"1) () 2  , FMicrosoft Equation 3.0 DS Equation Equation.39q<II  dArial e~(r"1~) FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfoqZEquation Native [X_1016263036wtFօjօjOle ]CompObjsu^fObjInfov`Equation Native a_1016263074yFjjl~IzI c d,d ( dArial e~(r"1~) ) 2 FMicrosoft Equation 3.0 DS Equation Equation.39qOle dCompObjxzefObjInfo{gEquation Native hpTHIoI ( dArial e~(r"1~) ) 2  FMicrosoft Word Picture MSWordDocWord.Picture.89q FMicrosoft Equation 3.0 DS Equation Equation.39q_1016277354 F&jI2jData _1Table}h5CompObjjh&DdB * S A? 23*:7]t/olDn`!d3*:7]t/o~L xt2xcdd``gd``beV dX,XĐ  A?d@e@P5< %! `f: `W01d++&10ԁ\ F\!" FP7f_  md L,6100@L  W&00|1p F]h6ʟȴlF e &2Y`%2pA  )wLLJ% V[!DdB  S A? 20_]~T-gjn`!_0_]~T-Ȉ@8 xt-xRJ@}3iS BA7c7)p~2#/"B&3@MfqHO2Hv؍:n"dLn`8A@wLLJ% ] V[1DdtB  S A? 2vWcXUb=Pwn`!ovWcXUb=P߮  xt=xJPƿs6- 8!BOP 8G *lAuwQpqvwrS {OoCkp=~9juBcUcmj1x>md <B  # = <B  # < r  6; <B  # : <B  # 9 r  68 <B  # 7 <B  # 6 x   <  <B ! # 5 <B "@ # 4 x # <# <B $ # 3 <B %@ # 2 <B ' # 1 <B (@ # 0 x ) < )  <B * # /<B +@ # .<B - # -<B .@ # ,x / < /  <B 0 # +<B 1@ # *<B 3 # )<B 4@ # (x 5 <5 <B 6 # '<B 7@ # &<B 9 # %<B :@ # $x ; <; <B < # #<B =@ # "<B ? # !<B @@ #  x A <A <B B # <B C # <B D # <B E # ~ F NBTCx DE|F1<Yew6 :mgM     V U67s&bl=p?Tx ?@@ G ZBTCw DE|F1;Jeh6w m:M g   )  U6U5l4=#lz  Tw ?@@ H ZBTC DE|F1Khe6XmM  8 e   U76FslC=~A T ?@@~ I NBTCr DE|F1KZe6;mM H  (   EU6QBl=z   Tr ?@@ J ZBTCDE|F1-Ze6g WWmM   c  _ 2U_6j  l =   T?@@ K ZBTCDE|F1w<e6gIvm*M  8    U6C/ l =  p"T?@@x M <M  x N <N  x O < O  x P < P  x Q < Q  NB R S D1TB S c $D1TB T c $D1x U <U  x V <V <2 W # k NB X S D <2 Y # k NB Z@ S DB S  ?  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOQDGtTTtOM tSTtPAtRT  tVKQ tZCVtYYTtXO TStWVJtUYtNtMWt ~tVt#t).t/ 4 t5, t;2tA!tK;tJ;tI;1tH ; tG-; tF ; tE;<tDtC;tB;t@;t?#t=;t<#t:;t9#t7;t6#t4;t3#t1 ; t0 # t. ; t- # t+;t*#t(;t'#t%;t$#t";t!#t;<t;H<tttHtfttHt0[ttHt tT U tT HU t t  t H t t Ht t ;t ;<tt;t;t;tQss{}  08DFSUV_`ikqrvw @t @GTimes New Roman5Symbol3& ArialG" HelveticaArial"1hDD&0dDavorka PetrinovicDavorka PetrinovicObjInfo~lObjectPool`*jI2j_1016276398F`*j`*jOle mCompObjnfObjInfopEquation Native qd_1016276481F`*j`*jHkIvI (c d,d   d2 ) FMicrosoft Equation 3.0 DS Equation Equation.39qXl~II (c d,dOle sCompObjtfObjInfovEquation Native wt   d(r"1) ) FMicrosoft Equation 3.0 DS Equation Equation.39q,0nI,I (c d,d "_1016276817F`*j`*jOle yCompObjzfObjInfo|Equation Native }H_1016276844F`*j`*jOle CompObjf FMicrosoft Equation 3.0 DS Equation Equation.39qToII ( d(r"1)  ) 2  ) FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfoEquation Native p_1016276968F`*jI2jOle CompObjfObjInfoEquation Native x_1016277333FI2jI2j\l~II ( ( d(r"1)  ) 2  ) FMicrosoft Equation 3.0 DS Equation Equation.39qHl~II ( d(rOle CompObjfObjInfoEquation Native d"1) ) 2Oh+'0l   ( 4 @LT\dssDavorka Petrinovicavo Normal.dotrDavorka Petrinovic2voMicrosoft Word 8.0@Ik@g@p[}gWordDocumentSummaryInformation(DocumentSummaryInformation81Table%Y bjbjWW ==Q2]d244440002222222$b3V5rC2!0Y'd 000C224"42220T&4L2022222 ( Fg22 50 55 60 65 70 75 80 4.88 min (ad2 ) and min( EMBED Equation.3 ) 4.90 min EMBED Equation.3  and min EMBED Equation.3  EMBED Equation.3  4.92 optimal crit., (11) and (13) 4.94 # zero elements [%] 4.96 Gp [dB] 4.98 re-optimized predictor zeroed full predictor :<>@BDFNPR\^䤛yj&EHUmH)j%< B*CJOJQJUVhmHnH6mHjEHUmH)jU)< B*CJOJQJUVhmHnH jUmH6CJH*mHCJH*OJQJmH6H*OJQJmH6OJQJmHmHCJB*CJOJQJhmH nH  jUmH/DFPR"#($x$:<>@BDFPR"#()=>CDLMRSjk7!")8<=DEFKLSijk¹ 6CJmH6CJH*mHCJmH6mHjNEHUmH)jl'< B*CJOJQJUVhmHnHjGEHUmH)jQ'< B*CJOJQJUVhmHnHmH jUmH()=>CDLMRSjk$N N!"#$%՜.+,D՜.+,4 hp   FER-ZESOI-j  Title 6> _PID_GUIDAN{DD55D0ED-0941-11D4-9D72-0040054232E7}n`!_Ђ%­o .pGH8xMhSAg%M׏/kj@K"TFT#V[@Uij^j$>jLT}ouOOw0{(7<բDpֳ;gKxg<\ӍVv5*/EÜrj_&W8!ݾ+ڣF{j{.TvW) &}&)J<&3cyŒqpίk<[͸[^Cq/tCQcTUS]"3kuJ\YkEE9=ofIon9="g@#>vv$5 =yo w-w/ŴM3TuDdhB D S A6? 52Iaڧ́tb %qn`!aڧ́tb @|xcdd``> @bD"L1JE `x,56~) M @ k!} UXRYvoav@Hfnj_jBP~nbC#%@y m˜v&7| P0pL@&91pA}%ة`f``GI)$53>:dDdB E S A7? 62ڴeqN ^:1]Pn`!UڴeqN ^:1h@C#xAJPWM FAWBQ9@X0* z\q"W}8o|>K  ChKc [b51VkɩǍU6!w$NI2|&Pg&ɪDcTҫ*L[ wetcC.de ]q6Qf0 )#lѷ#z! O*/Uhyr2w, >IBHN&]$<ۥqFUuj L4}DyK _Ref4793530537DdB G S A9? 82f.w{\M}n`!uf.w{\M@CCxAJP֦ (].*TЅ!`$;w vx%MA&SS:b"y.CږJɨ?6Uz\g*J,PѴpL.:|Q>!9Z""f|;/ Yx>Ep݁g k/:Dƙ Ǝ0uIt1Ȟ4w}O*tMGLE}J̝~O2%/)n }DyK _Ref4792527419DdB I S A:? 92 J5_wTM.rn`!w J5_wTM.*hFBExMLA߼mA*|ʶ B#FЃQ5AYPp@ F/AHbH4D`"51GA3] o7;mxa@ ~{Q\كtZgerc&.'<"&* - `}~EZm `/@s?"+"Wgj8GMqUl_W 2~ZY"#~@ [DPˍ_ƙZauZƱUXÎIT%@"Lj&tf^A\X'tr߅7JQr̰Bqi|dMN4VP얫]׷Tҕbxc/} L`/_f0B4–r[/]TOf)~GiH v= %amQLi'cǗs=,}1Q#n"-|ruuNB_[XZN*킿:a ^u8uSvt.]ӸiCBCxѳ;mIOH{ŏbO3]obJ~BoV{q1==3 1OyXp{=T4n1!|gRY;)&&7vi-^Lm!MWl3OAt3.P 'E5%iIa){O\ *DdXB J S A;? :2! ٥&uan`!Y! ٥&u~xt'xJ@ƿ_ӂAPS TxD|=ؼ@ĀƖ 7Y1<|x\gRb00o!sNcY˖$H;ZѶvZL~3]?0c։J]H"16L.)/U&e&s?Z}DyK _Ref479391681}DyK _Ref479391681}DyK _Ref479353053YDdOB K S A<? ;2l8B( sU=n`!l8B( sU0, xtexJ@ƿ Ƃ"QB0jJcK[\J=yzO3xVy5?bd73B^`A? 2wZL$($O)3+N`>G=uT~quܰMq/1> ]֯׎^`sfiP_LYj.lO,ӂ{mp}r2baU 1(0Uub}DyK _Ref479391681,0Dd!!B q S A'? =2/VgHYڙ$ǔxr/On`!j/VgHYڙ$ǔxu Lؒ:8/xZxն{i{KBJAD$PFR!#HAQB&AHWRwQ)6PUPϜ&@JXyN=O~d'?rlVPt@e8Ty#f%@g?"gzǬt#眙؊7Z'Zf..=3o=3*#[VKEVP=k@ُ9ș<ݻ9gfe*AKxLxZ>333\XB`pXʗRl3B`)_=GU-='W|5<_{^ُD,zfeBDLL43#%I(wScοŞ󯮱G_rRr8l!J`SSeT֎fk%/Ʋb)Dc%{Xm a$Vm>ժZvg~,W8x4 H;d}Nj P?RUMW_w7k vŽ/2G/@Wx Adɠ`Fc4~^a+h@dx\g` Sv}Ya1xH"`\0?b!d6g@f>XCݵ[+%XvvuXc;d{#~*K4>Cv5>b]5>ȧlPw=G%_C Vn`da_Z5+a~lYieͱ&&.`:AXIj3[of8H~+Uouj,oXC|>?zVjoB'H;+?.oVk~\ݍ{ {d|=M9|bog5^߲|.~%ox 9:2A3.p l'9r<s7N%Z锢NZ5=L9IT\֣8y{֔SlH%m i$RCي4X!2e_(rL x:(GN9 'Hڅ5>U4 _j>NTNͣj1ݨ`}2 ^HQ1(.TxJ}!3ԍb(VRWxF)RVO5T`=ZӪL5BB]rS*UTq'8q#*n+c El7E;Ott?C]ȀpKsܢ(ᮉUjܽ8n#G{q̭o|5gxM 0jyuo 1xÍ^JT&o1ܘ3"g#Zod;+ܞ0bEKQwVz(+ӓ,8oXVB>m*=S8w_wԀz+>D ;s^e?ď﷨I(F} +jz11Fl:+ &b~x*Cm jV9ԫ+f$fjⵡ_5.Yaw~7@u~t75 jX+ u-;G]*a[{.X/= >{XiQ#,CdžWj6 C|xMk<͇_K:XV_o ;FU~?gb3Bx[ D ψm7!.LsM/blfQK2y7Hec[ -A u_]Q54>a!V?77m ]/~?<D?,gxY0A XG^j>&4K5|Y\|Yu|YZ+Ͳ|Y1+ufyo5+2= oVfM~ȼ9y3?j&߻-v1k3y?oѸa%?;V ˜d X 6%ʠ߫ j'a൭a5u_FOekͱ6V[mЇMqOh:vVGz\ݠ{Z)"V{>Y`؄0|$ݽt}lOd lf՛/%V߃J/z~|V/?{aZOֳ++Mfi5h=ߠ{z_wN׸qyAEc,neѼz챼*jx-5쉼=n=^ߞC~= Ğ[oVifOvz{ ~dE%{ jh荲|"4MWY*>~G /<^ o ={|78i=eo[1sչ͜u S@@ڗzq> 8;x\ 4^:;lxUgyv 4~|Y:[nh [3>l;{x@vkGC ن[T<^?!|>/bA>MO똻fb{3,t8R6!z+c|so½|÷:滝#̿p3?ʏ9/}cỲ1AtG,DJ/$ P>Yh 婂,t"ՐըIwȪW֠djY~25njROm0sv鹮|:c&y9&S|/t8=6Cd)+1OiDe Mm5>Kv-\e #{*ً`~o{gY7 so g0 B'p/;%39CT!GƑFh<Mʤ"jǽXI5U1OOxe5*aξ-j@/Iͥ;167T-zH-Gp?ěCQWP:eZAObY_QZMq/`O3]\}@CvUj Wi&4>Y}D MU_Lu@Muh:zWuhV5רsQIYVv%O3e}*BG㪐8 UZR5URV/]8 ({p4^"{(&bnuvk8ƽKwox%N+݆{k#nomgfU63=sf(YEXX:]ZC@@jAPrARJiNiP@@TDA:$$Q޽wβ{ywf~g\]uIQ]Һ)]RI7nB*떤nC7u{DGmtgRw'mGf7nz0[>@"c= Ip-Gaz"'z*Ǔ1'nB&dG/s<||Wd^FVUdoJQ'-d^g>y$^rF"?=^;>F~ӧ$ϓ?eH\$諄D+^;5wIb'y'#A4y赳qhi6܆Z{OhID Z$62&)-nRҒ&-m~7 y$$-)V0iU6,orZ&3h#f-)NrlU4TmLmTCۛ4ʼK{1hwӊ6M{DAaX>M?: L˧!tI'1tl^h;cY۝s>$=Γi6tCY 8u @Wpb8t?ep=Q %`*%$%UcA!U=~Fzezf-jWw{2pe|,|>/:-2j-5-M9/sar p' ׃3C"I@8JNL$JIFkA@]Iz!\ IPw=I֓huR&%~X=Z=&P}Q>psRTTA +]zүS]Sإch'o9zgPZ擣N_X@iԆ!ܞ=D1T0FLX"<£X(^zf!Nǂ{Y|Ɇ%qB^`u`p^Uq))p}>.b8z<#yYGVCwrL^'ZQʖy^"zBRzOd; Uuhp5fQi*՝&WhigDOM-vFm:Lq:Ui6Kqf?Q)PUN&;*ۭ*}< qUgUciݞ>?rn;Y/aܕ/rwy^T~apG݂ q-&ɓ< K<Lu+^FM<g[^:_C-c4ôõtYHct+t}D#b'I/bqA9,n}}ZK<\ -eWp-Wh*h\\}]wpitWB' Oس \]eЯCp>5!E$'y-O@"P"Rֳu۳O <)2ҙDIqBg,F¤,*E*īgURU"QMRՁͶ ) e'{u PgR.P.O[>@ (Hs4A i=X;/Igo -%]Zm ,m'}N%2k]~&#/ij U2"SO2=!ХSP:%=+$B|Js2х(=Y2.G9 exYb`@A>kT<+QUV>FTXumLǢp5tזt8j׺ ۖ}?,h(Fӑh,dT7 y͆ttNBі*FN'mt&M砽39Jgà @N}9vi//u@ۇt zD˖";<)[^cP*2:l%V[<>:U/[ zTfQ!{8TmCgiJll};~;lykv5gg{jnG=by_vdwQ?mW bF dY>F_@%%Koh\[l ųpѷP2!F=h?;O;S<=3xp܏gŃyNx\GLT_Ƃ'q1.~xHAL2b+~`(w'ųL1$˗B[*>ǫ,7/*oog.b.6=³n-> ~w; Y^G |"q—mS\[m#,%& D$$DRr{J+5I%0d-L25!ӑjyAFLNɌ7 \\F$ߴO",Iޖ-o,Ki"9gX/ҳ~(kβ!ۖ"~9$[acdK2R%e2Y|lG(2[v#N/7Z#eO{]+FpfC`e&r4NN$;gm+'t͇gm=&?'<}ʧ"eʍL.ɯr'!Y~G&r[#ʃ?=y-/Gu(u;;D?IIX$WOHE-|4rijM#J$4JN3gShh.fy^TiA UhUQ%h)Ur-Ѳ**Z^]UUmZM5uT=ڪZ)}[U ˛6jG.Y[U7z6?>j mЎjFXK]D`i5j6[LUj9utZmhU[HǩmOR{h1:U|:I'3.埫+t.]nYXI?WCkeEJKV)cmkU[K6P˷l,^e|zmQyaT!kO٫J4ۡel*v2yU>U 7`U]OFjNjaEՖVueN_S%՛]QU_Al FZf4ˑ;+/-WRݕ60㮱<);B,Oa~baӺ'Y ,K^fLUֽ˲3w{euf9]uY^%<+xA7/*k)&E^ }uJn^`y7ay Ҽ[nY̭#}ڼ6o6]p(%}Vp^n'`@C| >0>Gxw)|; \;|>R>/ܕ|h{^A=—s|3|ߺ7z6;G; 0.&"I,M!.Nauӊnf_Z~710}O,_<ieKZ<{.`Gg(ϭCEp/:DSb:(X^DsQJ&~buTZ?#^UxM}տi^_q^O˛ꃼś7[xk=_ozm2AQkN/ǻͻwy=xo=~zGaz D<3X#u>Fw|nw7Q~;õN:Dټ'@Bc!)wOCD=x>K|ħl=/3,_?B`b˗|^ l_YNo6Z˷|> $ߦYC 5Cc1߯FI>rFKqL`IHmӗu2qV$ "n: +&oYP" cSCԦafHiF'(3I"|&29"aynbF1-/`fk6fVQl[2?Z}9O7D sD QA6*(k΋ 抨n.[^\Uz CQ<u 6w"[$2)&T67eLfiL`rYeL!`%処=LYT]MuT#5w,d˾&R0Ph!e[>#ɉfŞr#')r*l+4 p7>H{C՝Ͼ!T=|wN'|m|Vp?V<77iR|$f)lq2%=Ձ˝aR|(%Qd&N)T^8 ?&\:6΋tΣ6wݥGŨ}xXVQZkվU]Sy:Pd uAv"⼷ҫA'ILmelUr8 /Tq^a3 <>:F;vڶ}va-v͕#,KDDְ"7k~5̯t~7/SЯQ155D ]5/@d }䩞־mgk\})cv Ya8!#p˵ꇰ/ql-\נĭP(Mit&$ O ;۽tj*P{l_j?ԁ8Ʃּk=;rNW_pNg{Iǫݘ/YV'E|0YemwY L2Eԗ(O !39Et lNc9Gpt]Ưu멜/:=۷ϯqV`/~{luzvڪ=c) >~S"bAߔ$z<|X.\U83\WpI*ї.8?9c '^6_G}y%CϣO `~g8驑+:gWO _y<ŧW+Qfg%:eW,+Jx(L9⿯DcDYƺOD;R處&RUFwH̒+^KNl/X3?WF99EcwU"ө] WWݻ*9YF]e8/ڻ65_;>"b]òؽ~9g^m}%*И'N@?]$>X {GbG*y6 b#1d<{c#1s ĕ %u~:F.Y'K$x_m_sxtK˿="/sOx{l_<~/"k~]/'d.'d~xR'?Xvۇ"ZoE|M}Et> C`'D|C)T׌v{oDulo?rn>,Fjol۫UdXVm[v)T0jS:E_o >ŵ{kOL7I"}\ Nc2}#_>z }ƠkӁS_:ETNn?ؽ1C!ߴf~ 'R~-O) /!ԏ_G]ԑ?DMyB\gexq\7 G0#kp>~\8'p%~y5?;h! _PID_GUIDAN{69B280C9-F40B-11D3-9D72-0040054232E7} [$@$NormalmH H@!2H Heading 1$$$x@&5;CJKHF@F Heading 2$x<@&56CJOJQJ<A@<Default Paragraph Font@>@@Title$<@&5CJ KHOJQJ0O!0abstract<6mH "O!2"equ $(8B@28 Body Text $1$<CJmH HM@!2HBody Text First Indent <F"@FCaption$!;x @CJmH 0OR0Slika$$$xmH 0O120Figura$s:O!r: Reference & Fe<.@. Footnote Text8&@8Footnote ReferenceH*DW]Hi ]H  ]H !     5 Z%p4AGNUM-P-U?CLV     / 1 J ^ ` ')3GI^rt-19MOfz|Qeg`tv  [ o q !!!!!######$$$$$$(%<%>%A%S%V%`%t%v%(((-(0(D(F(v***+++...>>*>>>@>X>p>u>%B=B?B]H: t: t: t:: t:t: t:: t:::t::t: ttt:::: t:::t:: t: ttt: t:t:tttt:::::::::::::::::t:::::::: t:::t:t: t:ttt::t: O"$@ӻT'^ld$2$\9cw؅!2$Ez"Z#4LM)@0(  B S  ?]H= _Ref479157793 _Ref478283845 _Hlt475952499 _Ref479148100 _Ref479145595 _Ref479252741 _Hlt479326489 _Ref479147624 _Ref479148754 _Ref479151316 _Ref479155808 _1015155867 _1015156914 _1015157583 _1015158314 _1015159527 _1015159621 _1015159938 _1015164221 _1015164403 _1016184937 _1016192159 _1016192535 _1016193100 _1016193849 _1016198936 _1016204609 _1016206698 _1016219917 _1016223131 _1016223154 _1016223206 _1016223245 _1016223628 _Ref479353053 _Hlt479336499 _Ref479391681 _1016260590 _1016260676 _1016260762 _1016261256 _1016261765 _1016261859 _1016262000 _1016263124 _1016264633 _1016276058 _1016276149 _1016276535 _1016277055 _1016277091 _1016277118 _1016277146 _1016277152 _1016277378 _Ref465842149 _Hlt478280496 _Ref465842182 _Hlt479140118 _Ref465842264 _Ref465842297 F !#@%55555555555555555555555#9q;;=B=B=B=B=B=B=B=B=B=B=B=B=B=B=B=B=B=BEEEEEF^H @ @ @@@@@@@@@@@@@@@@@@@ @!@"#$%@&@'@(@)@*@+@,@-@.@/@0@1@2@3@4@5@6@87;9:<3 i +!#X%55555555555555555555555;9;;=B=B=B=B=B=B=B=B=B=B=B=B=B=B=B=B=B=BEEE.FFG^H 2 J a *^u2 !!##$$$$(%?%A%W%`%w%(.(v**++..v>%B@BGRH^H##$$$$A%W%(.(v**++..v>GGGGGGPHPHRH[H^HDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka3D:\users\davorka\radovi\Pula2000_Sven\Novi_rad3.docDavorka Petrinovic3C:\Users\Davorka\radovi\Pula2000_Sven\Novi_rad4.docA -V&"r0]$-o(.@hh[]o(.V&"0]$A   @$$ad$$l#l$%&*+a/a00F0G]H @,@*X@.`@4l@:x@R@GTimes New Roman5Symbol3& Arial"0hrCD&D&q # 9{0dFG%SPARSE VECTOR LINEAR PREDICTION WITH Davorka PetrinovicDavorka Petrinovic  FMicrosoft Word Document MSWordDocWord.Document.89q