arxiv: v1 [] 24 Feb PDF

Decodng Generazed Reed-Soomon Codes and Its Appcaton to RLCE Encrypton Schemes arxv: v1 [] 24 Feb 2017 Yongge Wang Department of SIS, UNC Charotte, USA. February 28,
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Decodng Generazed Reed-Soomon Codes and Its Appcaton to RLCE Encrypton Schemes arxv: v1 [] 24 Feb 2017 Yongge Wang Department of SIS, UNC Charotte, USA. February 28, 2017 Abstract Ths paper presents a survey on generazed Reed-Soomon codes and varous decodng agorthms: Berekamp-Massey decodng agorthms; Berekamp-Wech decodng agorthms; Eucdean decodng agorthms; dscrete Fourer decodng agorthms, Chen s search agorthm, and Forney s agorthm. Key words: Reed-Soomon code; generazed Reed-Soomon code. 1 Fnte feds 1.1 Representaton of eements n fnte feds In ths secton, we present a Layman s gude to severa representatons of eements n a fnte fed GF(q). We assume that the reader s famar wth the fnte fed GF(p)=Z p for a prme number p and we concentrate on the constructon of fnte feds GF(p m ). Poynomas: Let π(x) be an rreducbe poynoma of degree m over GF(p). Then the set of a poynomas n x of degree m 1 and coeffcents from GF(p) form the fnte fed GF(p m ) where fed eements addton and mutpcaton are defned as poynoma addton and mutpcaton moduoπ(x) For an rreducbe poynoma f (x) GF(p)[x] of degree m, f (x) has a rootαn GF(p m ). Furthermore, a roots of f (x) are gven by the m dstnct eementsα,α p,,α pm 1 GF(p m ). Generator and prmtve poynoma: A prmtve poynoma π(x) of degree m over GF(p) s an rreducbe poynoma that has a rootαn GF(p m ) so that GF(p m )={0} {α : =0,, p m 1}. As an exampe for GF(2 3 ), x 3 + x+1 s a prmtve poynoma wth rootα=010. That s, α 0 = 001 α 1 = 010 α 2 = 100 α 3 = 011 α 4 = 110 α 5 = 111 α 6 = 101 α 7 = 001 Note that not a rreducbe poynomas are prmtve. For exampe 1+ x+ x 2 + x 3 + x 4 s rreducbe over GF(2) but not prmtve. The root of a generator poynoma s caed a prmtve eement. Matrx approach: The companon matrx of a poynomaπ(x)=a 0 + a 1 x+ +a m 1 x m 1 + x m s defned to be the m m matrx M= a 0 a 1 a 2 a m 1 1 The set of matrces 0, M,, M pm 1 wth matrx addton and mutpcaton over GF(p) forms the fnte fed GF(p m ). Spttng fed: Letπ(x) GF(p)[x] be a degree m rreducbe poynoma. Then GF(p m ) can be consdered as a spttng fed ofπ(x) over GF(p). That s, assume thatπ(x)=(x α 1 ) (x α m ) n GF(p m ). Then GF(p m ) s obtaned by adjonng these agebrac eementsα 1,,α m to GF(p). 1.2 Fnte fed arthmetcs Let α be a prmtve eement n GF(q). Then for each non-zero x GF(q), there exsts a 0 y q 2 such that x=α y where y s caed the dscrete ogarthm of x. When fed eements are represented usng ther dscrete ogarthms, mutpcaton and dvson are effcent snce they are reduced to nteger addton and subtracton moduo q 1. For addtons, one may use Zech s ogarthm whch s defned as Z(y) : y og α (1+α y ). (1) That s, for a fed eementα y, we haveα Z(y) = 1+α y. If one stores Zech s ogarthm n a tabe as pars (y, Z(y)), then the addton coud be cacuated as α y 1 +α y 2 =α y 1 (1+α y 2 y 1 )=α y 1 α Z(y 2 y 1 ) =α y 1+Z(y 2 y 1 ). For the fnte fed GF(2 m ), the addton s the effcent XOR operaton. Thus t s better to store two tabes to speed up the mutpcaton: dscrete ogarthm tabe and exponentaton tabes. For the dscrete ogarthm tabe, one obtans y on nput x such that x=α y. For the exponentaton tabe, one obtans y on nput x such that y=α x. In order to mutpy two fed eements x 1, x 2, one frst gets ther dscrete ogarthms y 1, y 2 respectvey. Then one cacuates y=y 1 + y 2. Next one ooks up the exponentaton tabe to fnd out the vaue ofα y. Note that we have x 1 x 2 =α y 1 α y 2 =α y 1+y 2. 2 Reed-Soomon codes 2.1 The orgna approach Let k n q and a 0,,a n 1 be dstnct eements from GF(q). The Reed-Soomon code s defned as C= { (m(a 0 ),,m(a n 1 )) : m(x) s a poynoma over GF(q) of degree k }. There are two ways to encode k-eement messages wthn Reed-Soomon codes. In the orgna approach, the coeffcents of the poynoma m(x)=m 0 + m 1 x+ +m k 1 x k 1 s consdered as the message symbos. That s, the generator matrx G s defned as 1 1 a 0 a n 1 G=..... a k 1 0 a k 1 n 1 and the the codeword for the message symbos (m 0,,m k 1 ) s (m 0,,m k 1 )G. Letαbe a prmtve eement of GF(q) and a =α. Then t s observed that Reed-Soomon code s cycc when n=q 1. For each j 0, et m=(m 0,,m k 1 ) and m = (m 0 α 0, m 1 α 1,,m k 1 α k 1 ). Then m (α )=m 0 α 0 + m 1 α 1 α + +m k 1 α k 1 α (k 1) = m(α +1 ). That s, m s encoded as ( m (α 0 ),,m (α n 1 ) ) = ( m(α),,m(α n 1 ), m(α 0 ) ) 2 whch s a cycc shft of the codeword for m. Instead of usng coeffcents to encode messages, one may use m(a 0 ),,m(a k 1 ) to encode the message symbos. Ths s a systematc encodng approach and one can encode a message vector usng Lagrange nterpoaton. 2.2 The BCH approach We frst gve a defnton for the t-error-correctng BCH codes of dstanceδ. Let 1 δ n=q 1 and et g(x) be a poynoma over GF(q) such that g(α b )=g(α b+1 )= =g(α b+δ 2 )=0 whereαs a prmtve n-th root of unty (note that t s not requred to haveα GF(q)). It s straghtforward to check that g(x) s a factor of x n 1. For w=n deg(g) 1, a message poynoma m(x)=m 0 + m 1 x+ +m w x w over GF(q) s encoded as a degree n 1 poynoma c(x)=m(x)g(x). A BCH codes wth b=1 s caed a narrow-sense BCH code. A BCH code wth n=q m 1 s caed a prmtve BCH code where m s the mutpcatve order of q moduo n. That s, m s the east nteger so thatα GF(q m ). A BCH code wth n=q 1 andα GF(q) s caed a Reed-Soomon code. Specfcay, et 1 k n=q 1 and et g(x)=(x α b )(x α b+1 ) (x α b+n k 1 )=g 0 + g 1 x+ +g n k x n k be a poynoma over GF(q). Then a message poynoma m(x)=m 0 + m 1 x+ +m k 1 x k 1 s encoded as a degree n 1 poynoma c(x) = m(x)g(x). In other words, the Reed-Soomon code s the cycc code generated by the poynoma g(x). The generator matrx for ths defnton s as foows: g 0 g 1 g n k 0 0 g(x) 0 g 0 g n k 1 g n k 0 xg(x) G= = g n 2k+1 g n 2k+2 g n k x k 1 g(x) For BCH systematc encodng, we frst choose the coeffcents of the k argest monomas of c(x) as the message symbos. Then we set the remanng coeffcents of c(x) n such a way that g(x) dvdes c(x). Specfcay, et c r (x)=m(x) x n k mod g(x) whch has degree n k 1. Then c(x)=m(x) x n k c r (x) s a systematc encodng of m(x). The code poynoma c(x) can be computed by smuatng a LFSR wth degree n k where the feedback tape contans the coeffcents of g(x). 2.3 The equvaence The equvaence of the two defntons for Reed-Soomon code coud be estabshed usng dscrete Fourer transform (Mattson-Soomon poynoma). In the foowng, we assume that n=q 1 andαs a prmtve eement of GF(q). For a poynoma f GF(q)[x] wth deg( f ) n, the Mattson-Soomon poynoma of f s defned as n Φ( f )(x)=f(x)= f (α )x n. It s straghtforward thatφs a near map over GF(q)[x]. By the fact that x n 1=(x 1)(1+ x+ + x n 1 ), we have x(x n 1)=(x 1)(x+ x x n ). Thus n a = 0 for a a GF(q) wth a 1. Let f (x)= n 1 j=0 a jx j. Then F(α j ) = n f (α )α j(n ) = n n 1 u=0 a uα u α j(n ) = n 1 u=0 a n u α (u j) (2) = na j 3 If f (x) s a mutpe of the generatng poynoma g(x)= ( ) n k j=1 x α j, then f (α j )=0 for 1 j n k. Thus F(x)=Φ( f ) has degree at most k 1. The dentty (2) gves us an nverse ofφas foows: Φ 1 (F)(x)=n 1 Φ(F)(x 1 ) mod x n 1. Φ 1 maps poynoma of degree at most k 1 to mutpes of the generatng poynoma g(x). For each Reed-Soomon codeword f (x) n the BCH approach, t s a mutpe of the generatng poynoma. The above arguments show that F(x) has degree at most k 1 and (F(α 0 )/n,,f(α n 1 )/n)= 1 n f (x). Thus f (x) s aso a Reed-Soomon codeword n the orgna approach. For each Reed-Soomon codeword (a 0,,a n 1 ) n the orgna approach, t s an evauaton of a poynomas F(x)=Φ( f ) of degree at most k 1 onα 0,,α n 1. By the dentty (2), (a 0,,a n 1 ) s the coeffcents ofφ 1 (F) whch s a mutpe of the generatng poynoma g(x). 2.4 Generazed Reed-Soomon codes For an [n, k] generator matrx G for a Reed-Soomon code, we can seect n random eements v 0,,v n 1 GF(q) and defne a new generator matrx v v 1 0 G(v 0,,v n 1 )=G..... = G dag(v 0,,v n 1 ) v n 1 The code generated by G(v 0,,v n 1 ) s caed a generazed Reed-Soomon code. For a generazed Reed- Soomon codeword c, t s straghtforward that c dag ( v 1 0 n 1),,v 1 s a Reed-Soomon codeword. Thus the probem of decodng generazed Reed-Soomon codes coud be easy reduced to the probem of decodng Reed-Soomon codes. 3 Decodng Reed-Soomon code 3.1 Peterson-Gorensten-Zerer decoder Ths sectons descrbes Peterson-Gorensten-Zerer decoder whch has computatona compexty O(n 3 ). We assume that the Reed-Soomon encodng process uses the BCH s systematc encodng process. We frst cacuate syndromes. Let r(x)=c(x)+e(x) where e(x)= e p x p has t n k 2 non-zero coeffcents. The numbers 0 p 1,, p t n 1 are error postons and e p are error magntudes (vaues). For convenence, we w use X =α p to denote error ocatons and Y = e p to denote error magntudes. The syndromes S j for j=1,,n k are defned as S j = r(α j )=c(α j )+e(α j )=e(α j )= e p (α j ) p = Y X j. 4 That s, we have X1 1 X 1 X 1 t Y 1 S 1 X1 2 X2 2 Xt 2 Y 2 S 2. = (3) X1 n k X2 n k Xt n k Y t S n k Thus we obtaned n k equatons wth n k unknowns: X 1,, X t, Y 1,,Y t. The error ocator poynoma s defned as t Λ(x)= (1 X x)=1+λ 1 x+ +λ t x t. (4) Then we have Λ(X 1 )=1+λ 1 X 1 + +λ t X t = 0 (,,t) (5) Mutpy both sdes of (5) by Y X j+t, we get Y X j+t Λ(X 1 )=Y X j+t +λ 1 Y X j+t 1 + +λ t Y X= j 0 (6) For,,t, add equatons (6) together, we obtan (Y X j+t )+λ 1 (Y X j+t 1 )+ +λ t (Y X j )=0 (7) Combng (3) and (7), we obtan S j λ t + S j+1 λ t 1 + +S j+t 1 λ 1 + S j+t = 0 ( j=1,,t) (8) whch yeds the foowng near equaton system: S 1 S 2 S t λ t S t+1 S 2 S 3 S t+1 λ t 1 S t =... S t S t+1 S 2t 1 λ 1 S 2t (9) Snce the number of errors s unknown, Peterson-Gorensten-Zerer tres varous t from the maxmum n k 2 to sove the equaton system (9). After the error ocator poynoma Λ(x) s dentfed, one can use exhaustve search agorthm (or Chen s search agorthm for hardware mpementatons) to fnd the roots of Λ(x). After the error ocatons are dentfed, one can use Forney s agorthm to determned the error vaues. Wth e(x) n hand, one substracts e(x) from r(x) to obtan c(x). Computatona compexty: Assume that (α j ) for =0,,n 1 and j=0,,n k have been precomputed n a tabe. Then t takes 2(n 1)(n k) fed operatons to compute the vaues of S 1,,S n k. Affter S are computed, t takes O(t 3 ) fed operatons (for Gaussan emnatons) to sove the equaton (9) for each chosen t Chen s search agorthm The probem of fnd roots for the error ocator poynoma n (4) coud be soved by an exhaustve search. Aternatvey, one may use Chen s search whch s based on the foowng observaton. Λ(α ) Λ(α +1 ) = 1+λ 1 α + +λ t (α ) t = 1+λ 1, + +λ t, = 1+λ 1 α λ t (α +1 ) t = 1+λ 1, α+ +λ t, α = 1+λ 1,+1 + +λ t,+1 5 Thus, t s suffcent for us to compute the set{λ j, :,,q 1; j=1,,t} wthλ j,+1 =λ j, α j. Chen s agorthm can be used to mprove performance for hardware mpemenaton though t has mted advantage for software mpementaton wth ogarthm and exponentaton tabe ook-up for fed mutpcatons Forney s agorthm For Forney s agorthm, we defne the error evauator poynoma (note that n k 2t) and the syndrome poynoma Note that S (x)λ(x) Ω(x)=Λ(x)+ X Y x t j=1, j S (x)=s 1 x+s 2 x 2 + S 2t x 2t. (1 X j x) (10) = ( 2t t =1 Y X x) tj=1 (1 X j x) mod x 2t+1 = t Y 2t =1 (X x) t j=1 (1 X j x) mod x 2t+1 = t Y (1 X x) (11) 2t =1 (X x) t j=1, j (1 X j x) mod x 2t+1 Usng the fact that (1 x 2t+1 )=(1 x)(1+ x+ + x 2t ), we have Thus Ths gves us the key equaton 2t (1 X x) (X x) = X x (X x) 2t+1 = X x mod x 2t+1. =1 S (x)λ(x)= Y X x t j=1, j (1 X j x) mod x 2t+1. Ω(x)=(1+S (x))λ(x) mod x 2t+1. (12) Note: In some terature, syndrome poynoma s defned as S (x)=s 1 + S 2 x++ S 2t x 2t 1. In ths case, the key equaton becomes Ω(x)=S (x)λ(x) mod x 2t. (13) LetΛ (x)= t X j (1 X j x)= t λ x 1. Then we haveλ (X 1 )= X j (1 X j X 1 ). By substtutng X 1 ntoω(x), we get Ths shows that Ω(X 1 )= X Y X 1 t j=1, j (1 X j X 1 )=Y t j=1, j ) e p = Y = X Ω(X 1 Λ (X 1 ) (1 X j X 1. )= Y X 1 Λ (X 1 ) Computatona compexty: Assume that (α j ) for =0,,n 1 and j=0,,n k have been precomputed n a tabe. Furthermore, assume that both Λ(x) and S (x) have been cacuated aready. Then t takes O(n 2 ) fed operatons to cacuateω(x). After bothω(x) andλ(x) are cacuated, t takes O(n) fed operatons to cacuate each e p. As a summary, assumng that S (x) andλ(x) are known, t takes O(n 2 ) fed operatons to cacuate a error vaues. 6 3.2 Berekamp-Massey decoder In ths secton we dscuss Berekamp-Massey decoder [1] whch has computatona compexty O(n 2 ). Note that there exsts an mpementaton usng Fast Fourer Transform that runs n tme O(n og n). Berekamp- Massey agorthm s an aternatve approach to fnd the mnma degree t and the error ocator poynoma Λ(x)=1+λ 1 x +λ t x t such that a equatons n (8) hod. The equatons n (8) defne a genera near feedback shft regster (LFSR) wth nta state S 1,,S t. Thus the probem of fndng the error ocator poynomaλ(x) s equvaent to cacuatng the near compexty (aternatvey, the connecton poynoma of the mnma ength LFSR) of the sequence S 1,,S 2t. The Berekamp-Massey agorthm constructs an LFSR that produces the entre sequence S 1,,S 2t by successvey modfyng an exstng LFSR to produce ncreasngy onger sequences. The agorthm starts wth an LFSR that produces S 1 and then checks whether ths LFSR can produce S 1 S 2. If the answer s yes, then no modfcaton s necessary. Otherwse, the agorthm revses the LFSR n such a way that t can produce S 1 S 2. The agorthm runs n 2t teratons where the th teraton computes the near compexty and connecton poynoma for the sequence S 1,,S. The foowng s the orgna LFSR Synthess Agorthm from Massey [1]. 1. Λ(x)=1, B(x)=1, u=1, L=0, b=1, =0. 2. If = 2t, stop. Otherwse, compute L d=s + λ j S j (14) j=1 3. If d=0, then u=u+1, and go to (6). 4. If d 0 and 2L, then Λ(x)=Λ(x) db 1 x u B(x) u=u+1 and go to (6). 5. If d 0 and 2L, then T(x)=Λ(x) Λ(x)=Λ(x) db 1 x u B(x) L=+1 L B(x)=T(x) b=d u=1 6. =+1 and go to step (2). Dscusson: For the sequence S 1,,S, we use L = L(S 1,,S ) to denote ts near compexty. We use Λ () (x)=1+λ () 1 x+λ() 2 x2 + +λ () L x L to denote the connecton poynoma for the sequence S 1 S that we have obtaned at teraton. At teraton, the constructed LFSR can produce the sequence S 1 S 2 S. That s, L S j = λ () j S j, j= L + 1,, =1 Let 0 denote the ast poston where the near compexty changes durng the teraton and et d denote the dscrepancy obtaned at teraton usng the equaton (14). That s, L 1 d = S + λ ( 1) j S j. j=1 7 (15) We show thatλ () (x)=λ ( 1) (x) d b 1 x u B(x) s the connecton poynoma for the sequence S 1,,S. The case for d = 0 s trva. Assume that d 0. Then B(x)=Λ ( 0) (x) and b=d By the constructon n Step 4 and Step 5, we haveλ () (x)=λ ( 1) (x) d d xu Λ ( 0) (x). For v= L, L + 1,, 1, we have S v + L j=1 λ() j S v j = S v + ( L 1 j=1 λ( 1) j S v j + d d S v L 0 { 0 L u 1 = d d d d 0 +1 u= j=1 λ( 0) j S v j Computatona compexty: As we have mentoned n Secton 3, t takes 2(n 1)(n k) fed operatons to cacuates the sequence S 1,,S n k. In the Berekamp-Massey decodng process, teraton requres at most 2( 1) fed operatons to cacuate d and at most 2( 1) operatons to cacuate the poynomaλ () (x). Thus t takes at most 4t(2t 1) operatons to fnsh the teraton process. In a summary, Berekamp-Massey decodng process requres at most 2(n 1)(n k) + 4t(2t 1) fed operatons. 3.3 Berekamp-Wech decoder Ths secton dscusses Berekamp-Wech decodng agorthm whch has computatona compexty O(n 3 ). It s noted that Berekamp-Wech decodng agorthm frst appeared n the US Patent 4,633,470 (1983). The agorthm s based on the cassca defnton of Reed-Soomon codes and can be easy adapted to the BCH defnton of Reed-Soomon codes. The decodng probem for the cassca Reed-Soomon codes s descrbed as foows: We have a poynoma m(x) of degree at most k 1 and we receved a poynoma c(x) whch s gven by ts evauatons (r 0,,r n 1 ) on n dstnct fed eements. We know that m(x)=r(x) for at east n t ponts. We want to recover m(x) from r(x) effcenty. Berekamp-Wech decodng agorthm s based on the fundamenta vanshng emma for poynomas: If m(x) s a poynoma of degree at most d and m(x) vanshes at d+ 1 dstnct ponts, then m s the zero poynoma. Let the graph of r(x) be the set of q ponts: {(x, y) GF(q) : y=r(x)}. Let R(x, y) = Q(x) E(x)y be a non-zero owest-degree poynoma that vanshes on the graph of r(x). That s, Q(x) E(x)r(x) s the zero poynoma. In the foowng, we frst show that E(x) has degree at most t and Q(x) has degree at most k+t 1. Let x 1,, x t be the st of a postons that r(x ) m(x ) for,,t where t t. Let E 0 (x)=(x x 1 )(x x 2 ) (x x t ) and Q 0 (x)=m(x)e 0 (x). By defnton, we have deg(e 0 (x))=t t and deg(q 0 (x))=t + k 1 t+k 1. Next we show that Q 0 (x) E 0 (x)r(x) s the zero poynoma. For each x GF(q), we dstngush two cases. For the frst case, assume that m(x)=r(x). Then Q 0 (x)=m(x)e 0 (x)=r(x)e 0 (x). For the second case, assume that m(x) r(x). Then E 0 (x) = 0. Thus we have Q 0 (x) = m(x)e 0 (x) = 0 = r(x)e 0 (x). Ths shows that there s a poynoma E(x) of degree at most t and a poynoma Q(x) of degree at most k+t 1 such that R(x, y) = Q(x) E(x)y vanshes on the graph of r(x). The arguments n the precedng paragraph shows that, for the mnma degree poynoma R(x, y) = Q(x) E(x)y, both Q(x) and m(x)e(x) are poynomas of degree at most k + t 1. Thus Q(x) m(x)e(x) has degree at most k+t 1. For each x such that m(x) r(x)=0, we have Q(x) m(x)e(x)=0. Snce m(x) r(x) vanshes on at east n t postons and n t k+t 1, the poynoma R(x, m(x))= Q(x) m(x)e(x) must be the zero poynoma. 8 ) The equaton Q(x) E(x)r(x)=0 s caed the key equaton for the decodng agorthm. The arguments n the precedng paragraphs show that for any soutons Q(x) of degree at most k+t 1 and E(x) of degree at most t, Q(x) m(x)e(x) s the zero poynoma. That s, m(x)= Q(x) E(x). Ths mpes that, after sovng the key equaton, we can cacuate the message poynoma m(x). Let (m(a 0 ),,m(a n 1 )) be the transmtted code and (r 0,,r n 1 ) be the receved vector. Defne two poynomas wth unknown coeffcents: Usng the denttes Q(x)=u 0 + u 1 x+ +u k+t 1 x k+t 1 E(x)=v 0 + v 1 x+ +v t x t Q(a )=r E(a ) (=0,,n 1) to bud a near equaton system of n equatons n n+ 1 unknowns u 0,,u k+t 1, v 0,,v t. Fnd a non-zero souton of ths equaton system and obtan the poynoma Q(x) and E(x). Then m(x)= Q(x) E(x). Computatona compexty: The Berekamp-Wech decodng process soves an equaton system of n equatons n n+1 unknowns. Thus the computatona compexty s O(n 3 ). 3.4 Eucdean decoder Assume that the poynoma S (x) s known aready. By the key equaton (12), we have Ω(x)=(1+S (x))λ(x) mod x 2t+1 wth deg(ω(x)) deg(λ(x)) t. The generazed Eucdean agorthm coud be used to fnd a sequence of poynomas R 1 (x),,r u (x), Q 1 (x),,q u (x) such that x 2t+1 Q 1 (x)(1+s (x))=r 1 (x) 1+S (x) Q 2 (x)r 1 (x)=r 2 (x) R u 2 (x) Q u (x)r u 1 (x)=r u (x) where deg(1+s (x)) deg(r 1 (x)), deg(r (x)) deg(r +1 (x)) (,,u 1), deg(r u 1 (x)) t, and deg(r u (x)) t. By substtutng frst u 1 denttes nto the ast dentty, we obtan the key equaton Λ(x)(1+S (x)) Γ(x)x 2t+1 =Ω(x) where R u (x)=ω(x). In case that the syndrome poynoma s defned as S (x)=s 1 + S 2 x++ S 2t x 2t 1, the Eucdean decoder w cacute the key equaton Λ(x)S (x) Γ(x)x 2t =Ω(x) Computatona compexty: As we mentoned n the prevous sectons, t takes 2(n 1)(n k) fed operatons to cacuate the poynoma S (x). After S (x) s obtaned, the above process stops n u steps where u t+1. For each dentty, t requres at most O(t) steps to obtan the par of poynomas (R, Q ). Thus the tota steps requred by the Eucdean decoder s bounded by O(t 2 ). 3.5 Dscrete Fourer transform decoder The dscrete Fourer transform maps a poynoma v(x)=v 0 + v 1 x+ +v n 1 x n 1 to ts vaues DFT(v 0, v 1,,v n 1 )=(v(a 1 ),,v(a n )) 9 where a 1,,a n are fxed dstnct ponts from GF(q). DFT s a bjecton on GF(q) n wth the nterpoaton as the nverse transform, whch s often caed the nverse dscrete Fourer transform DFT 1. Let c(x) be the code word and r(x) = c(x)+ e(x) be the receved word. The dscrete Fourer transforms of c(x), e(x), and r(x) coud be defned as C(x), E(x), and R(x) resp
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks