Professional Documents
Culture Documents
Lecture 715 PDF
Lecture 715 PDF
1. Discrete-time Wiener process. Let Zn , n ≥ 0 be a discrete time white Gaussian noise (WGN)
process, i.e., Z1 , Z2 , . . . are i.i.d. ∼ N (0, 1). Define the process Xn , n ≥ 1 as X0 = 0, and
Xn = Xn−1 + Zn for n ≥ 1.
(a) Is Xn an independent increment process? Justify your answer.
(b) Is Xn a Markov process? Justify your answer.
(c) Is Xn a Gaussian process? Justify your answer.
(d) Find the mean and autocorrelation functions of Xn .
(e) Specify the first and second order pdfs of Xn .
(f) Specify the joint pdf of X1 , X2 , and X3 .
(g) Find E(X20 |X1 , X2 , . . . , X10 ).
Solution:
(a) Yes. The increments Xn1 , Xn2 − Xn1 , . . . , Xnk1 − Xnk are sums of different Zi random
variables, and the Zi are IID.
(b) Yes. Since the process has independent increments, it is Markov.
(c) Yes. Any set of samples of Xn , n ≥ 1 are obtained by a linear transformation of IID
N (0, 1) random variables and therefore all nth order distributions of Xn are jointly
Gaussian (it is not sufficient to show that the random variable Xn is Gaussian for each
n).
(d) We have " n n n
#
X X X
E[Xn ] = E Zi = E[Zi ] = 0 = 0,
i=1 i=1 i=1
1
(e) As shown above, Xn is Gaussian with mean zero and variance
Substituting, we get
X1 0 1 1 1
X2 ∼ N 0 , 1 2 2 .
X3 0 1 2 3
2. Arrow of time. Let X0 be a Gaussian random variable with zero mean and unit variance,
and Xn = αXn−1 + Zn for n ≥ 1, where α is a fixed constant with |α| < 1 and Z1 , Z2 , . . . are
i.i.d. ∼ N (0, 1 − α2 ), independent of X0 .
2
Solution:
(a) Yes, the process is Gaussian, since it is the linear transform of white Gaussian process
{Zn }.
(b) Yes, the process is Markov since Xn |{X1 = x1 , . . . , Xn−1 = xn−1 } ∼ N (αxn−1 , 1 − α2 ),
which depends only on xn−1 .
(c) First note that RX (n, n) = α2 RX (n − 1, n − 1) + (1 − α2 ) = 1 for all n. Since we can
express
Xn = αk Xn−k + αk−1 Zn−k+1 + · · · + Zn ,
and Xn−k is independent of (Zn−k+1 , . . . , Zn ), we have RX (n, n−k) = E(Xn Xn−k ) = αk .
Thus, RX (n, m) = α|n−m| .
(d) Because the process is Gaussian, the MMSE estimator is linear. From Markovity, X̂100 =
E(X100 |X1 , . . . , X99 ) = E(X100 |X99 ) = RRXX(100,99)
(99,99) X99 = αX99 .
(e) Again from the Markovity and the symmetry, (X1 , . . . , Xn ) has the same distribution as
(Xn , . . . , X1 ) hence X̂100 = E(X100 |X101 , . . . , X199 ) = E(X100 |X101 ) = αX101 .
(f) First note that X100 is conditionally independent of (X1 , . . . , X98 , X102 , . . . , X199 ) given
(X99 , X101 ). To see this, consider
Thus,
3
where the amplitude A(t) is a zero-mean WSS process with autocorrelation function RA (τ ) =
1
e− 2 |τ | , the phase Θ is a Unif[0, 2π) random variable, and A(t) and Θ are independent. Is
X(t) a WSS process? Justify your answer.
Solution: X(t) is wide-sense stationary if EX(t) is independent of t and if RX (t1 , t2 ) depends
only on t1 − t2 . Consider
and
4
4. LTI system with WSS process input. Let Y (t) = h(t) ∗ X(t) and Z(t) = X(t) − Y (t) as shown
in the Figure 1.
(a) Find SZ (f ).
(b) Find E(Z 2 (t)).
Your answers should be in terms of SX (f ) and the transfer function H(f ) = F[h(t)].
X(t) Y(t)
h(t)
+ -
Z(t)
Solution:
RZ (τ ) = E(Z(t)Z(t + τ ))
= E((X(t) − Y (t))(X(t + τ ) − Y (t + τ )))
= RX (τ ) + RY (τ ) − RY X (−τ ) − RXY (−τ )
= RX (τ ) + RY (τ ) − RXY (τ ) − RXY (−τ ).
(b) To find the average power of Z(t), we find the area under SZ (f )
Z ∞
2
E(Z (t)) = |1 − H(f )|2 SX (f ) df.
−∞
5. Echo filtering. A signal X(t) and its echo arrive at the receiver as Y (t) = X(t) + X(t − ∆) +
Z(t). Here the signal X(t) is a zero-mean WSS process with power spectral density SX (f ) and
the noise Z(t) is a zero-mean WSS with power spectral density SZ (f ) = N0 /2, uncorrelated
with X(t).
Solution:
5
(a) We can write Y (t) = g(t) ∗ X(t) + Z(t) where g(t) = δ(t) + δ(t − ∆).
Thus, SY (f ) = |G(f )|2 SX (f ) + SZ (f ) = |1 + e−j2π∆f |2 SX (f ) + N20 .
(b) Since SY X (f ) = (1 + e−j2π∆f )SX (f ),
6. Discrete-time LTI system with white noise input. Let {Xn : −∞ < n < ∞} be a discrete-time
white noise process, i.e., E(Xn ) = 0, −∞ < n < ∞, and
(
1 n = 0,
RX (n) =
0 otherwise.
The process is filtered using a linear time invariant system with impulse response
α n = 0,
h(n) = β n = 1,
0 otherwise.
The impulse response h(n) is the sequence (α, β, 0, 0, . . .). The convolution with h(−n) has
only finitely many nonzero terms.
7. Finding time of flight. Finding the distance to an object is often done by sending a signal
and measuring the time of flight, the time it takes for the signal to return (assuming speed of
signal, e.g., light, is known). Let X(t) be the signal sent and Y (t) = X(t − δ) + Z(t) be the
6
RY X (t)
t
2 5 8
Figure 2: Crosscorrelation function.
signal received, where δ is the unknown time of flight. Assume that X(t) and Z(t) (the sensor
noise) are uncorrelated zero mean WSS processes. The estimated crosscorrelation function of
Y (t) and X(t), RY X (t) is shown in Figure 2. Find the time of flight δ.
Solution: Consider
Now since the maximum of |RX (α)| is achieved for α = 0, by inspection of the given RY X we
get that 5 − ∆ = 0. Thus ∆ = 5.
7
Additional Exercises
Do not turn in solutions to these problems.
(a) Let Xn = 12 Zn−1 + Zn for n ≥ 1. Find the mean and autocorrelation function of Xn .
(b) Is {Xn } wide-sense stationary? Justify your answer.
(c) Is {Xn } Gaussian? Justify your answer.
(d) Is {Xn } strict-sense stationary? Justify your answer.
(e) Find E(X3 |X1 , X2 ).
(f) Find E(X3 |X2 ).
(g) Is {Xn } Markov? Justify your answer.
(h) Is {Xn } independent increment? Justify your answer.
(i) Let Yn = Zn−1 + Zn for n ≥ 1. Find the mean and autocorrelation function of Yn .
(j) Is {Yn } wide-sense stationary? Justify your answer.
(k) Is {Yn } Gaussian? Justify your answer.
(l) Is {Yn } strict-sense stationary? Justify your answer.
(m) Find E(Y3 |Y1 , Y2 ).
(n) Find E(Y3 |Y2 ).
(o) Is {Yn } Markov? Justify your answer.
(p) Is {Yn } independent increment? Justify your answer.
Solution:
(a)
1
E(Xn ) = E(Zn−1 ) + E(Zn ) = 0.
2
RX (m, n) = E(Xm Xn )
1 1
=E Zn−1 + Zn Zm−1 + Zm
2 2
1 2
2 E[Zn−1 ], n−m=1
1 E[Z 2 ] + E[Z 2 ], n = m
4 n−1 n
= 1 2 ],
2
E[Z n m−n=1
0, otherwise
5
4, n = m
1
=
2 , |n − m| = 1
0 , otherwise.
8
(b) Since the mean and autocorrelation functions are time-invariant, the process is WSS.
(c) Since (X1 , . . . , Xn ) is a linear transform of a GRV (Z0 , Z1 , . . . , Zn ), the process is Gaus-
sian.
(d) Since the process is WSS and Gaussian, it is SSS.
(e) Since the process is Gaussian, the conditional expectation (MMSE estimate) is linear.
Hence,
5 1 −1 X1
1 4 2
1
E(X3 |X1 , X2 ) = 0 2 1 5 = (10X2 − 4X1 ).
2 4 X 2 21
(f) Similarly, E(X3 |X2 ) = (2/5)X2 .
(g) Since E(X3 |X1 , X2 ) 6= E(X3 |X2 ), the process is not Markov.
(h) Since the process is not Markov, it is not independent increment.
(i) The mean function µn = E(Yn ) = 0. The autocorrelation function
2 n = m,
RY (n, m) = 1 n = m − 1 or n = m + 1,
0 otherwise.
(j) Since the mean and autocorrelation functions are time-invariant, the process is WSS.
(k) Since (Y1 , . . . , Yn ) is a linear transform of a GRV (Z0 , Z1 , . . . , Zn ), the process is Gaus-
sian.
(l) Since the process is WSS and Gaussian, it is SSS.
(m) Since the process is Gaussian, the conditional expectation (MMSE estimate) is linear.
Hence,
2 1 −1 Y1
1
E(Y3 |Y1 , Y2 ) = 0 1 = (2Y2 − Y1 ).
1 2 Y2 3
(n) Similarly, E(Y3 |Y2 ) = (1/2)Y2 .
(o) Since E(Y3 |Y1 , Y2 ) 6= E(Y3 |Y2 ), the process is not Markov.
(p) Since the process is not Markov, it is not independent increment.
3. Random binary waveform. In a digital communication channel the symbol “1” is represented
by the fixed duration rectangular pulse
1 for 0 ≤ t < 1
g(t) =
0 otherwise,
9
and the symbol “0” is represented by −g(t). The data transmitted over the channel is
represented by the random process
∞
X
X(t) = Ak g(t − k), for t ≥ 0,
k=0
where A0 , A1 , . . . are i.i.d random variables with
1
+1 w.p. 2
Ai = 1
−1 w.p. 2.
Solution:
(a) The first order pmf is
pX(t) (x) = P (X (t) = x)
∞
!
X
=P Ak g(t − k) = x
k=0
= P A⌊t⌋ = x
= P (A0 = x) IID
1
= 2 , x = ±1
0 , otherwise.
Now note that X(t1 ) and X(t2 ) are dependent only if t1 and t2 fall within the same time
interval. Otherwise, they are independent. Thus, the second order pmf is
pX(t1 )X(t2 ) (x, y) = P (X (t1 ) = x, X (t2 ) = y)
∞ ∞
!
X X
=P Ak g (t1 − k) = x, Ak g (t2 − k) = y
k=0 k=0
= P A⌊t1 ⌋ = x, A⌊t2 ⌋ = y
P(A0 = x, A0 = y), ⌊t1 ⌋ = ⌊t2 ⌋
=
P(A0 = x, A1 = y), otherwise
1
2 , ⌊t1 ⌋ = ⌊t2 ⌋ & (x, y) = (1, 1), (−1, −1)
= 1
, ⌊t1 ⌋ =
6 ⌊t2 ⌋ & (x, y) = (1, 1), (1, −1), (−1, 1), (−1, −1)
4
0 , otherwise.
(b) For t ≥ 0,
∞
" #
X
E[X(t)] = E Ak g(t − k)
k=0
∞
X
= g(t − k)E[Ak ]
k=0
= 0.
10
For the autocorrelation RX (t1 , t2 ), we note once again that only if t1 and t2 fall within
the same interval, will X(t1 ) be dependent on X(t2 ); if they do not fall in the same
interval then they are independent from one another. Then,
1
where {Zi } is an i.i.d. process with P(Z1 = −1) = P(Z1 = +1) = 2. Define the absolute
value random process Yn = |Xn |.
11
Solution:
(a) This is a straightforward calculation and we can use results from lecture notes. If k ≥ 0
then
P{Yn = k} = P{Xn = +k or Xn = −k} .
If k > 0 then P{Yn = k} = 2P{Xn = k}, while P{Yn = 0} = P{Xn = 0}. Thus
n
1 n−1
(n+k)/2 2 k > 0, n − k is even, n − k ≥ 0
1 n
P{Yn = k} = n
k = 0, n is even, n ≥ 0
n/2 2
0 otherwise.
(b) If Y20 = |X20 | = 0 then there are only two sample paths with max1≤i<20 |Xi | = 10 that
is, Z1 = Z2 = · · · = Z10 = +1, Z11 = · · · = Z20 = −1 or Z1 = Z2 = · · · = Z10 =
−1, Z11 = · · · = Z20 = +1. Since the total number of sample paths is 20
10 and all paths
are equally likely,
2 2 1
P max Yi = 10|Y20 = 0 = = = .
1≤i<20 20 184756 92378
10
5. Mixture of two WSS processes. Let X(t) and Y (t) be two zero-mean WSS processes with
autocorrelation functions RX (τ ) and RY (τ ), respectively. Define the process
Find the mean and autocorrelation functions for Z(t). Is Z(t) a WSS process? Justify your
answer.
Solution: To show that Z(t) is WSS, we show that its mean and autocorrelation functions
are time invariant. Consider
µZ (t) = E[Z(t)]
= E(Z|Z = X)P{Z = X} + E(Z|Z = Y )P{Z = Y }
1
= (µX + µY )
2
= 0,
and similarly
RZ (t + τ, t) = E[Z(t + τ )Z(t)]
1
= (RX (τ ) + RY (τ )) .
2
Since µZ (t) is independent of time and RZ (t + τ, t) depends only on τ , Z(t) is WSS.
12
6. Stationary Gauss-Markov process. Consider the following variation on the Gauss-Markov
process discussed in Lecture Notes 8:
X0 ∼ N(0, a)
Xn = 12 Xn−1 + Zn , n ≥ 1,
(a) Find a such that Xn is stationary. Find the mean and autocorrelation functions of Xn .
(b) (Optional.) Consider the sample mean Sn = n1 ni=1 Xi , n ≥ 1. Show that Sn converges
P
to the process mean in probability even though the sequence Xn is not i.i.d. (A stationary
process for which the sample mean converges to the process mean is called mean ergodic.)
Solution:
(a) We are asked to find a such that E(Xn ) is independent of n and RX (n1 , n2 ) depends
only on n1 − n2 . For Xn to be stationary, E(Xn2 ) must be independent of n. Thus
E(Xn2 ) = 41 E(Xn−1
2
) + E(Zn2 ) + E(Xn−1 Zn ) = 14 E(Xn2 ) + 1 .
Therefore, a = E(X02 ) = E(Xn2 ) = 43 . Using the method of lecture notes 8, we can easily
verify that E(Xn ) = 0 for every n and that
(b) To prove convergence in probability, we first prove convergence in mean square and then
use the fact that mean square convergence implies convergence in probability.
Xn n n
1 1X 1X
E(Sn ) = E Xi = E(Xi ) = 0 = 0.
n n n
i=1 i=1 i=1
Thus Sn converges to the process mean, even though the sequence is not i.i.d.
13
7. QAM random process. Consider the random process
X(t) = Z1 cos ωt + Z2 sin ωt , −∞ < t < ∞ ,
where Z1 and Z2 are i.i.d. discrete random variables such that pZi (+1) = pZi (−1) = 21 .
(a) Is X(t) wide-sense stationary? Justify your answer.
(b) Is X(t) strict-sense stationary? Justify your answer.
Solution:
(a) We first check the mean.
E(X(t)) = E(Z1 ) cos ωt + E(Z2 ) sin ωt = 0 · cos(ωt) + 0 · sin(ωt) = 0 .
The mean is independent of t. Next we consider the autocorrelation function.
E(X(t + τ )X(t)) = E((Z1 cos(ω(t + τ )) + Z2 sin(ω(t + τ )))(Z1 cos(ωt) + Z2 sin(ωt)))
= E(Z12 ) cos(ω(t + τ )) cos(ωt) + E(Z22 ) sin(ω(t + τ )) sin(ωt)
= cos(ω(t + τ )) cos(ωt) + sin(ω(t + τ )) sin(ωt)
= cos(ω(t + τ ) − ωt)) = cos ωτ .
The autocorrelation function is also time invariant. Therefore X(t) is WSS.
(b) Note that X(0) = Z1 cos 0 + Z2 sin 0 = Z1 , so X(0) has the same pmf as Z1 . On the
other hand,
X(π/4ω) = Z1 cos(π/4) + Z2 (sin π/4)
1
= √ (Z1 + Z2 )
2
1
√
x = ± √2 = 2
4
2
∼ 21 x = 0
0 otherwise
This shows that X(π/4ω) does not have the same pdf or even same range as X(0).
Therefore X(t) is not first-order stationary and as a result is not SSS.
8. Finding impulse response of LTI system. To find the impulse response h(t) of an LTI system
(e.g., a concert hall), i.e., to identify the system, white noise X(t), −∞ < t < ∞, is applied
to its input and the output Y (t) is measured. Given the input and output sample functions,
the crosscorrelation RY X (τ ) is estimated. Show how RY X (τ ) can be used to find h(t).
Solution: Since white noise has a flat psd, the crosspower spectral density of the input X(t)
and the output Y (t) is just the transfer function of the system scaled by the psd of the white
noise.
N0
SY X (f ) = H(f )SX (f ) = H(f )
2
N0
RY X (τ ) = F −1 (SY X (f )) = h(τ )
2
14
Thus to estimate the impulse response of a linear time invariant system, we apply white
noise to its input, estimate the crosscorrelation function of its input and output, and scale it
by 2/N0 .
15