This preview shows page 1. Sign up to view the full content.
Unformatted text preview: v irtue o f orthogonality; t hat is, all t erms of r cos nwot cos mwot dt lTD (3.96a) ITo s tands for integration over a ny c ontiguous interval of To seconds. By
where
using a trigonometric identity (see Sec. B.76), Eq. (3.96a) can be expressed as I=~[r c os(n+m)wotdt+ r c os(nm)wotdt]
2 lTD
lTD (3.96b) 3 Signal R epresentation by Orthogonal Sets 224 Since cos wot e xecutes one complete cycle during any interval of To d uration, cos (n+
m )wot executes (71 + m ) complete cycles during any interval of d uration To· Therefore, t he first integral in Eq. (3.96b), which represents t he a rea under (n + m)
complete cycles o f a sinusoid, equals zero. T he same argument shows t hat t he second integral in E q. (3.96b) is also zero, except when n = m. Hence I in Eq. (3.96)
is zero for all n f m. When n = m, t he first integral in Eq. (3.96b) is still zero, b ut
t he second integral yields r dt r cos nwot cos mwot dt ={ I =~ 2 l To = To
2 T hus ~ n fm {~ n fm l To T (3.97a) m =nfO Using similar arguments, we c an show t hat r sin nwot sin mwot dt
l~ = 2 (3.97b) n =m fO a nd r sin nwot cos mwot dt =0 (3.97c) for all n and m l To Appendix 3C: Orthogonality o f t he Exponential Signal S et
T he s et of exponentials
val of d uration To, t hat is, e inwot (n = 0, ± 1, ±2, . .. ) is orthogonal over any inter m fn
m (3.98) =n Let t he i ntegral o n t he lefthand side of Eq. (3.98) be 1.
(3.99)
T he case m = n is trivial. In this case t he i ntegrand is unity, a nd I = To· W hen
fn m I= 1
j (m  n)wo e j(mn)wot It'
tl 1
e j(mn)wotl [ ej(mn)woTo
J(m  n)wo +To = .  1] = 0 T he l ast result follows from t he fact t hat woTo = 271", a nd ej27rk = 1 for all integral
values of k. 3 .9 Summary T his c hapter discusses t he foundations of signal representation in terms of its
components. T here is a perfect analogy between vectors and signals; t he analogy is 3.9 S ummary 225 so strong t hat t he t erm ' analogy' understates t he reality. Signals a re n ot j ust l ike
vectors. Signals a re vectors. T he inner or scalar product of two (real) signals is t he
a rea u nder t he p roduct of t he two signals. I f t his inner or scalar p roduct is zero,
t he signals are said t o be orthogonal. A signal f (t) has a component c x(t), where
c is t he i nner p roduct of f (t) a nd x (t) divided by E x, t he energy of x (t).
A good measure of similarity of two signals f (t) a nd x (t) is t he correlation coefficient en, which is equal t o t he i nner p roduct of f (t) a nd x (t) divided by J EfE x .
I t c an be shown t hat  1 :S en :S 1. T he m aximum similarity (en = 1) occurs only
when t he two signals have t he s ame waveform within a (positive) multiplicative
constant, t hat is, when f (t) = K x(t). T he m aximum dissimilarity (en =  1) occurs only when f (t) =  Kx(t). Zero similarity ( en = 0) occurs when t he signals
are orthogonal. In binary communication, where we a re required to distinguish
between t he two known waveforms in t he presence of noise a nd d istortion, selecting t he two waveforms with maximum dissimilarity (en =  1) provides maximum
distinguishability.
J ust as a vector c an be represente...
View
Full
Document
 Spring '13
 Bayliss
 Signal Processing, The Land

Click to edit the document details