One Property of The Finite Length Uncorrelated
Sequences
(Received May 27,1982)
by Hiroshi KONDO and Hiroaki TAKAJHO
Ab8tru¢t
The interesting property of the finite length uncorrelated sequences often called white noise in the signal processing is presented;the center power(the squared magnitude at the origin)of the autocorrelation in time averaging sense ls e卯al t◎坊e marginal e況rgy(total energy except at the◎どigin)◎f it.
蕊蕊,eq。。ncei,。ft。nu、edindigじ 琢κ)一ぱ: {3>
tal signal processing due to its advantageous There exist another practical way to define property・However the uncorreiated sequence the mean and the autocorrelation, that is time generated by・adig輌tal c()mputer is necessarily()f averag輌ng meth◎d.
漂:窯隠隠1慧蕊蒜隠e㍑ 恥)〕一ξ理捲ω (4)
is, the uncorrelated sequence is windowed by a φηη(K)=Ez〔η(ゴ)κ(ゴ十κ)〕
「e
Gこa瓢麗㌶:蕊;li血_。fa ±ガ喜㌦(の・嫡κ)(5)signal and noise is known not accurately but a
P・i。・iby assumpti。n。, i, g。tt,n f,。m th,。b. 屹・e E・〔・〕d蜘…t輌m・・v・・ag・・
…v・ti・n by time av。・。gi。g i。 e,g。di、ity. ln .F°「品w・−dim…拍・al・e卿・・w・・a・・ひ th輌ρ,, w。 di、郷。。e。f th。 p,字ti。、。f、u。h lla「1y d・fi・・th・mean・nd aut・c・・rel・ti・n・f・・
・・・…e1・t・d・eq・・nces i・time av。・。ging sen,e. 9xample・th・tw・−dim・n・i・n・l versi・n・f Eq・(5)
1s expressed as
2Mniti・n and P・・P・・ty・甜n…典・d ψ (κ1)−E,偏∫).功+瓦ノ+1)〕
Thご㌫ll。t。d、e_ce、κωwith。e,。一聯..蹴冨㌦( ・ノ)・η(ゴ+κノ+1)
mean has a following pmperty. (6)
E〔%(の〕={) (i==◎,1,2,… ) In Practice IV and〃are not able to be輌nfin輌ty.
E〔呼)・η(ノ)〕一{ ;e「°c tant■〈1)蕊;。器1ご鑑㌶:灘蹴s
Tぱe罐〔・〕denotes an ensenbleautocorrelation ψηη(κ)a瓢)i,ψ一(頁)一捲(か嫡κ),
defined as (}≦κ≦A7−1 (7)
ψ。。(κ)−E〔功)・μ( +K)〕一。・4(κ)(2) whe・e,
whereαis a nonzero constant and 4κ)is a Kronecker s delta function:
鬼(2v→づ)=鬼(〆).
Generally the value obtained from Eq s(5)and 3. One Property of The Autocorrelation
黙猫蕊a;蒜麗蒜。:11ili。器: N・w w・will d・fine the aut・c・rrel・ti・n・f th・Eq,、(5)。nd(6).1。 Fig.(1)η(のi、 a whit,.unif。,m urc°「「elated seq竺ence as in三q・(5)・R・g・・di・g
ψ。。(κ)i,n。t。qual t。 ze,。 at th。 p。i。tκ≠。 ψ・・(!)except f・「・一゜ m・・glnal・nr・gy ・..
b。t。,cil。te ab,uptly. It s f°und that thr{・ll・wlng・el・t1…h1P・xlst
It is sh・w・th・tψ・・(κ)(κ〉・)i・an un・・rre−be ln these quantlt es
:蕊ぽe;hen NapP「°aches infinity A一ψ2。(・);cent・・p・w・・ (9)
1・・tead・f Eq・.(5) B−2誓;ψ三。(の;m・・ginal・n・・gy (10)
VVe can also deflneψπη(κ)as . . .
(1)one−dlmenslonal case
ψ・・(κ)一誌一1N喜1功)・( +κ).⑧ ⑧If・ωi・awhite−Gaussian・equence
二ぽ蒜ぷ゜1。蕊le。㍑se°f its 酬±1て祠1(・eeApP・ndi・2)(1D
⑤Ifκ(のis a white−uniform sequence
φ。。(々)告睾4{5両1(・ee ApP・ndi・2)
(12)
(ii)Two−dimensional case With
∠1=ψ元η(0,0);center power (13)
ハし ノ
B=Σ Σψ元η(κ,1);
κ=一(N−D, ;一(2V−1)
々 .
marglnal energy(except atκ=1=0)(14)
Fig.l The autocorrelation of lmoorrelated ⑧ Ifη( ,ノ)is a white−gaussian sequence (white・un置orm)8equence with N・{辺
Fig2 The autocorrelation of㎜correlated(white・ga㎎sian)
sequence with N=M=64(Two・dimension)
診≒発;圭i両1(・ee apP・ndi・3) 1隠蕊潔1 sq囎「eこwhe廊
⑤竃華ごiご。蒜3、wぽぷ∵(・{・・d輌x5){17)
f,器蕊蕊黙罐ご漂t・ご灘麟:rl蒜f d8
placlng N琉th 2V2. E〔ノ▲一ノT〕2→0,2V→○○(see Apρend輌x 6) {18>
Even if we define the autocorrelation as in Eq. E〔B−B〕2→0,2V→、oo(see ApPendix 7) (19)
va「1ance砺=°・1・The value ls「° nded・t th・ Th・・e p・・賦。,。,e、ati,fl。d wh。th。,娠)is
麗劉。㍑㌶認{1)蒜e、惣an°「m・1seq…ce・−if・一・eq・−
th。 F輌9雛e(3随at this v。lue is alm・st centα・f 5・Pr・bability D・nsity Functi・n
th・v・・i・ti… Wh・・碑)i、。n。,m。1、e呼。。。 w量th,α。
Thi・v・「i・ti・n dep・nd$up°n th・unc・「「elated @mea・,・a・iance。三, th。 p・。b。bility d,n、ity f。。,,
・・ss・f th・・eq・・n・・gen…t・d by a c°mp・t・・ ti。n。fAi,giv,na、
Here, this se卯ence is genαated in d◎ubleμeci・
議驚欝欝嚥議曇∫(A)−29・篇r(芸)幽・一σ(㊨
e「at・d by・・ing the c・mp・t・・lib「・・y・f品COM= @wh。,e r(・)i、 k。。w。。、 g。mm。 fun,ti。n.
M−2◎◎(・逗n・lp・ed・i・・)・ (、ee ApP。。di。8>
The.sa口e is als°t「u警if N is va「ied bgt the Th輌・d。。・ityノ(A)d《d、。紬。 tw。 P。,。.
va「1at °n ls la「ge c°nsplcu°usly when Nls less m。t,rs N。。dσ釜.
th・n128・ Th。、 th, d。。、ity。f A i, d。t。,mi。,d if。。e
20 kn・ws the mean・fψ。。(0)and the number N of
degrees◎f丘eed《)m,
since
13
@ E{ψ。。(◎)}−E{κ6+κぞ+一・+鳩.、}一鳩く2D 9「 @ ・ Fig.(4)is the curve◎f/(A)佃1V=64 andσ。= メ0.1.
5 層 Thi, cu,v, h。、 a maxim。m val。。 at _1_,1.001.031.06、09L121.1i1.18合 A+貴)2煽
聴3H醜…・f合(N−5・2) Si・・e醐一(・+貴)2・・4,
跳蒜跳㌫瓢I th・p・i・t・t whi・h this c・・v・i・m・xi…i・
agree with E〔A〕衙en N is large.
Theρτobab輌蹴y dens柱y funαion◎f B is not
4:蒜蕊lncePmpe「tyl輌an 謬=。e蕊蒜ぽぱ
It is shown that the random variable A defined ever, that the densityノ(B)is roughly similar to
by Eq.⑨c◎nverges知the rand◎m variab至e B that◎f〆(A)since 8輌s greater也an zαo and the
variance of B decreases accoding to the increase E〔ψ朋(の・ψηη( 十ん)〕
°f礼 @ 一E〔捲・(1)κ(1− )
ア(∠D
・認・(11)・(11−(ゴ+〃))
1 N−1N−1 . .
〕
il =N・E恩。濯。〔η(1)・(11)・(1−・)・(1一(・+々))〕
コ
1 ! where l is not equal to 1− since is greater than
− ・e…If 11,1,1− ・nd(1r(ゴ+〃))・・e diff・・ent
l l each other, then the above expectation becomes
| l
I | zero.
ロ
i i If 1=11 andん=0, then l
_ 、4
0 んA E〔ψ。。(の・ψ。。( +〃)〕−E〔ψ。。(の〕・
A・=(1一貴)顕=(1+そ!傷・ 一繕1E〔・・(1)。・(1一の〕
9.4 probability density function of A
一耀〔Eη2(1)・Eη・(1− )〕
航Summa「y
@ 一寿、(ノV− )傷fi。濫h誌ご゜器。二盟1;st灌qぽ「7。lf:1:If 1−11−( +〃)and 11−1− ・th・n
center power of the autocorrelation in time 1=1− 一(ゴ十〃)
averaging is equal to its marginal energy in the =1_2 _〃.
m㌫::1「:1::n㌫w。th。t th。,e va,i。nce, tend Th・・
to zero as 2V apProaches infinity・ Finally the 〃==−2i;
probability density function of the center power E〔ψηη( ).ψ朋( 十ん)〕=E〔ψηη(ゴ).ψ朋(_ )〕
has been derived. =E〔ψnη(の〕2 The finite length uncorrelated sequence is often ≠O used for a practical digital signal processing . . .
becau,e。f it, analytical。dvant。g。、. .Thls 1旦natu「al becauseψ・・(ゴ) s an even func−
1。,u。h a ca,e th。 p,。P。,ty di、cussed i。 thi・ t °n°f・・ . .
P。p。, m。y b, utilized. C°nsequenly f°「・>Oand・ヰ >0,
蕊1議き羅竃竃ii三E〔 )・一)〕一{㌣∵w迎
correlated one, and further also apPlied to test a
,and。m,eq。。nce di,ectly. H・nce we c・n・1・d・th・tψ・・(2)i・an un・・「「e This re、ult、 can b。。f、。urse,ed。,i。。d i。。n l・t・d・equence f・・ >0・
uncorrelated process(continuous case)・ ApPendix 2:Proof of the fact that A/B tends
ApP。ndi。1:C。rrel。t。dness。fψ (のf。・ >O t°1
With
Since
Eψ一(の一寿曙・(1)κ(1一の A全ψ・膓(・)全〔捲2(1)〕2
一講E〔。(1)〕E〔。(1− )〕一・ 一;・〔慧κ2(1)〕〔鴛・2(11)〕
it・aut・c・rrel・ti・n i・identical with th・v・・iance 一是〔2V−1Ση4(1Z=0)+顯1・・(1)・・(11)〕
of ψηη( ). (zキz1)
we obtain Hence
EAr壱〔2v−1Σ飽4(1z=◎)+雛勘・(1)・Eη・(1・)〕 with N.、 。.、
(飼1) B=Σψ;η(り一2Σψ莞η(の 臨ωis a n・rmal灘d・m se緋nce we kn。w 〒謬 ;1
th・t(3
@ ≡2署Eψ莞。ω
酬)〕;3・』・E〔カ2(の〕=・5 鵠{(N−1)+(N−2)+・・一+1}
一嘉〔32V66十1V(ノV−1)・ξ〕 F斑h。m。,ew。。a。,。y蹴 一(1+寿)・3・ (A2−1) 題:: f。,N→。。
,
While if功)is a uniformly di$tributed random
Since the vaτi銀ces◎f/1 and 8 tend t◎zぼo when
sequence,
s輌nce it i§kn()wn that 2V approaches infinity as shown in App孤dixes
酬)〕一§。3。。d E〔。・(1)〕_5 (8)こ=ec。n。i。dea,N。pP,。ach。、mfi。ity
(a>咋):n◎mlal se堺ence
《1+5C)・》・ (A2−2)E_(・+5を)・〔ξ
・・th・・th・・h・・d
@ Eガブ(1寸)・6=N−1一 1ψ三・(1)一〔諾κ(の匁(膓一1)〕2 AP却ix 3・Tw・一』・迦al…ewlth r壱〔」V−1Σ冗2(のη2(/−4z=1) . the si〔N.
+顯》(11)・(1一1)・(ら)・(左一1)〕 (A2−3)』1㌔:竃覧㌘∵の
・thus we get
醐。(1)r試冒E〔・・(1)κ・(1−1)〕 ψ元・(丸1)r壱{慧冒・(らノ)・(ゴー泓ゲ1)}2
+糠醐1・)・(11−・)%(力)姻〕〕 一嘉{鱒輪)・・(声の
「羨〔2v−1ΣEη2(14=1)・飽・(1−1)+・〕 +買鑓1冒妬1,ノ1)功1一鳥元1−1)
Simi㌔1σ6 (AZ4) ・蕊竃竃㌫の}・(A卸
㌶にま蒜 竺灘∵三裁曇1裁
゜ (ゼ1キゴ2,ゴ1キゴ2)
Eψ三・(N−1)→、σ6 ・%・ω1)%・(勧)〕・晶
一〔N・(N・+2)・鵬〕・寿、 ca・e with N・i・place・f礼
一(1+寿、)・2 (A年2)霊蒜nc::PP「°aches infinit脇
⑤・(の:unif・・m
@ A1+昔N、+2
以一〔N・・法+(N・−N・)魔〕晶 万一「扉一M−1
撫晶_)∴÷誓一睾…
Eq・(A3−2)and(A3−3)are of the same form as ApPendix 4:Using the data cyclically one dimensional case with 2V2 in place of∧た (one dimensional case)
On the otherhand the marginal energy B is .
ψ莞η(0)1s the same value as a noncyclic case, i.e.
ハド エ ノ ハノ エ
B=4顯ψ編(〃,1)+2ハピ(丸゜) Eψち(・)−N(N+2)峠一(1+是)魔 +2恩ψ』・(0・1) (,ee, A2.1)
where 2V is assumed even. and since
From Eq.(A3−1)whenん≠Oor 1キ0蹴の一晶幣忽 蹴》念1;烈‡
・Eη2( 一〃,ノー1)+EΣΣ】ΣΣ!η( 1,ノ1)
1
1=ぞ蕊㌶、了2=Z =N 傷 万
・・(ゴ1一ん, ノ゜1−1)・功・・ゐ)・・( ・一疏一1)} 一己
一嬬(2V一ん)(N−1)傷 :
Consequently
イ・+N、)・; ・一⊇
Th・・ 一⊇・州・冒(ノV−〃−1)
告器一豊;圭i,f・・1・・g・亙 一場・寿・(N(1w1)」(㌻1)一ル・)
EB場ψ編(ん)+・聯是・ 蕊隠i蒜yhavethe;1:;蒜認
+継 P,蒜、蕊嬬tend、t°(1::膿dξ6㌻;
If娠)is n◎rmal,
E(∠1−」B)2==−2E(ノ1−ZD(B一亘)
4=EA=2V十2 =−2(五4B一ノ1B).
BEB N−1 ・ we。b,e,v。th。t
蒜:蕊よ。id驚1、隠t㌫:rご蒜朋8−E(捲・(の)・・2鷺(繕㌦(1)
sequ孤ceκω・ @ 一 嘱膓+々))2
一嬬・曙〆(り)(曇;・・ω)
:㍑蕊認llll6y ・曇:〔N−々−1Ση(1z=1)η(1+ん)〕Cヨ㌦(勿)』〃)〕
㌫゜) sthesame asthen°ncyclic・ 一嬬娼顯箸1〃箆㌦・伽・(ノ)
・灘(の%(/十珍)鬼(勿)ヵ(郷十ゐ).
聴!ξ。 )−E{ノV−1/V−1ΣΣ%(ゑの麓(カー9,ふノ〃=ObO)}2・詰、H・・c_ly wh。。1−_d キノキ/キ触 一曙慧・・(〃,1)・・(ん一義1づ)・嬬4 n°nze「°tαm「emains;
一顯卿)・E輌1一ノ)・晶 蠕・螺竃1菖㌦2(鋤2(鋤2(1)・2(1+ん)
−N・討、 ・nd・t』・・e・・輌・wi・g t。 th。。。effi。i。m
r誌 る・wh・頑・PP・・ach・・i・鮒y・
where η(一 ,一ノ)=η(1V− ,ノV一ノ), Thus using the independence ofη(の we obtain
…潮蜘)一(」V2−1)N・峠 翻 縄鷲菖㌦2(ゼ)躍2(ノ)
輔∴:1…+嘉)写・ ∋慧嘉螢篇酬)
腓Eψん(0,0)一蜘『2)傷・晶 一縄買量螢欝∵)
Thi、 i、。fth_m。f。 。, a_。y。1輌。 ca、e. r㌣一2(」v・−1」V2)。。。.
=(増,
ApPe dix 5:C°nve「g・nce i・th・mean・qua・・whil・f・・m Rq,.(A2ヨ),(A2.2)and(A2.5)wh。n
sense /V apProaches infinity
E〔A一β〕2:曇竺元巴晶二駕ピ撒) 互一万一傷 wh,,e肋d吾。,e th, m。。n、。f A。。dβC°nsequently
醐一8)2ゴ1㌫芦) −E〔 」v−1/v−14呂昆ψ』・(みψ》・(の〕一(吾〉・
一・ −4贈冨{纏㍍(ξ1)蕗1一の}2
㍑。隠C霊。』c°nvαges t°Bin the ・{捲(Z1)η(/一の}2}㈲・
ApP,ndix 6:V。,iance。f A 一糾冒冨{〔曇1娠1)娠・一の〕
中醐一互〕・ 、 〔《v−1Σ娠292泣ε)η(⇒}{〔置η(/1)η(/一の〕
二躍講(痴、 ・〔曇1棚左一の〕}}(吾)・
−E〔詰、(難・(の)4〕一④・ 一岩E〔雛〔{慧ηw(冷一の 一詞冨・・ω+4雛が(ノ)・・(〃) +鷲鯨( 一の・(あ)功・一の}
+3顯〆・(1)。・(;:鶴懲・(κ) ・{鐵W(一の+曇1鐵鮎 ・…・__㌍… 、・・(1・一/)η(/・)施一/)}〕〕一僻
忽瀞暴o弄〆( )霊ξ鷲一一 )
がど If %( )is n◎rmal then(3) ∧r_1」v_1
+Σ.Σ、%(ら)κ(ξ1−一)%( 2)%(ゴ2一の
一島 3 (η一1)唆1= .竃嶽)_)+鐵・(ん)。・(ん一 ).
Th酎h・無f…t・・m・a・ed・・P・・ω・et・ .冒冒。(11)。(1一/)。(1、)。(/、一の
び・f・・t・・素wh・廊・駆・a・h・・雄輌・ 莞欝
H…e +蕊忍、κω功1一鋤(22)如・一の
《ゼ1≒匡2)
㎡「鴇灘。鴛E・・ω・Eκ・(ノ1)・ ・顯・(11)・(11−1)・(1・)・(1・−1))}(吾)・
(∠1キゴ1キ々1キど1) (ξ1キど2)
・飽2㈲・飽2ω一(互)2 . . .
Thus the flrst term五s only remamed when 2V r亡{N4−2V−4(N2−」V)−3(N・−N) 。pP・。ach・・i・fi・ity.
一一6(N3−1V−3(N2−」V))}σξ一(ノZr)2
「亡恒N・+11N・−9N}場一④・1 ..、 ・』「岩E顯顯㌦W(ん一の
一。霧一④・. 』 ◆獅)・2(励〉一(万)2
Since互t・nd・t・σ2(・ee, A2−1)wh四N・p・ A, i。。pP。。di。6。nly wh,。バ1』キ〃翔th。
P・・ach・・i・f輌・ity・w・g・t f輌斑αM・・d・t・鴫・thαs…疏・h・d,・・d 媛一〇 f。,N−→。。 since告場(N→・・),
we have The identical result is also derived wh孤κ(の
i・・垣・if・・m・ σ』一→0, f。, N_。。.
Appendix 7 :variance of 8
。』−E〔B一吾〕・−E〔8・仁(君)・ ApPendix 8・P・・b・bility d・n・ity f…ti・・
ofA
−E〔(2誓ψ綱2〕一(君)・ S江e
A一ψ三・(・)一〔鑑2(1)〕2 ¥llle cisthec°efficientinEq・(A8−3N
wh_(。)一;詐(°)ア A(穿一り・・rヨ)+A一剖・1
↓=O Hence
Ifwe
@kn°wthep・d・£』・(°))°f R・・(°), then
@ A−1一撫.
(N)
綱)−1ゐ。(品。(0)) Ad・. .
ば 1力頗 尺・元(・)−M ρ脚・x9・Thef・Mhm限e輌fw曲・蝸
+1吉1椰・・(・))輪(。)。.晒 dist「ibuted「and°m va「iable
wh・・e W−∂R慧。) t,㌶慧ごin蕊↑溺織㍑;:
B・輪(&。(0))−Of・・凪。(0)<0,h・nce va「iance・ξi・ ・
砲)
瓢:)一 嬬:雀:ll三
P。霊蒜惣蒜篇惣a㌫霊i蒜,Th・・th・f・u・thm・m飽ti・
va
gジ
鋼一∫:蜘)血R 。(・)一冨・・(1) 一∬κ・2}μ h…chi−Sq・a・e・t・ti,ti。, w輌舳。。輌r ヨτ・
N鴛ご ル2)(一)y− :i弦
・〆 鬼《鰍2確)・σ(乱。(◎))
(A8.2) A・kn・w1・dg・m・nt
Thus from Eq.(A8コ) The authors wish to thank M.S. Matama for
姻一2一帯(M2)A㌍1 hishelpful advise i…°mpute「p「°⌒ing・
・ε一w砲σξ・σ(A) (A8−3) Refer磯ees
This densityム(A)depends・n the tw・para・ (1)Nahi, N E:Estirnati・n The・ry and ApPli・
meters 2V and 〈㌃. cat輌◎n, Losangeles, Caガforn輌a.
Thus the density of A is determined if one knows (January 1969)
the mean of Rηη(0) (2)Hino, M.:Spectral Analysis, Asakura shoten
;E{R。。(0)}−E{。6+一・・蹄.、}一鳩 (3)P・p卯li・・A・・P・・b・bihty, Rand・m陥・i・・
・nd th・加mb。・N(deg,e,、。f f,,ed。m). ble§・and St・…tl・P・・・・・…pぼ47・1就・・
Th・加・・f A・t w桓。h m。xim。m p。加。f nat1°nal st・d・nt・diti…M・GRAW−HILL Eq.〈A8−3)i。。bt。i。。d is s。。h臓 KOGAKUSHA・
禦一・((璽一梱・レ瀬・2鰯 (告場† ごブ⇒