精选习题 Q&A 20240630
精选习题 Q&A 20240630
精选习题 Q&A 20240630
这些习题建议大家抽时间做一下。并不会作为作业要求大家交,但是对巩固知识点有较
大作用。
填空选择题
1. The random variable 𝑋 follows the exponential distribution with parameter λ= 2,then
𝐸(𝑋) = ____________________, 𝐷(𝑋) = ____________________。
1 1
2
; 4
2. If the output of a system with a white noise is as follows, then the noise equivalent pass band
𝛥𝑓𝑒 is _____________________
L
2π
9
3. The power spectrum of stochastic processes 𝑋(𝑡) is 𝑆𝑋 (𝜔) = ,then the mean of it is
𝜔2 +9
_____________,variance is ___________.
3 3
𝑅𝑋 (𝜏) = 2 𝑒 −3|𝜏| , 𝑅𝑋 (∞) = 𝑚𝑋2 = 0, 𝜎𝑋2 = 𝑅𝑋 (0) − 𝑚𝑋2 = 2
4. Assume that there is a speech source ‘a’, located at A, and one microphone ‘b’ and
microphone located at B. The sound velocity (声速) is 340m/s, and the distance between A
and B is d(AB)=17m. The source ‘a’ send out a speech signal from time _______ to
________(second). Assume that the microphone ‘b’ sample the signal under a sampling rate
𝑋 from time 0 to 0.1s (second), and what microphone ‘b’ get is [0,0,0,0,0,1,2,3,0,0,0] where
the ′1,2,3′ is the signal part. Then 𝑋=_________Hz.
0 to 0.03s
100Hz
引申:如果出现多个麦克风阵列?如果是远场的情况?例如:麦克风 1 和 2 差 1 个 sample,
那么 1 和 3?如果 1 和 3 差 1 个 sample,那么 1 和 7? 如果麦克风 1 能收到全部信号但麦
克风 N 无法收到全部信号?
5. For the random walk model, the probability of ‘+1’ is 𝑝 = 0.3. We define the probability of
the event ‘return to the origin at stage 2𝑛’ as 𝑢2𝑛 , and ‘first return to the origin at stage 2n’
as 𝑣2𝑛 .
a) the probability of ‘return to the origin sooner or later’ is ____________;
b) It is known that the event ‘return to the origin stage 10’, defined as event ‘A’, happened,
then the probability of ‘first return to the origin at stage 16’ under event ‘A’ happened is
__________;
Sol:
a) 𝑉(𝑧 = 1) = ∑∞ 𝑛=0 𝑣2𝑛 𝑧
2𝑛
= 1 − √1 − 4𝑝𝑞𝑧 2 = 1 − √1 − 4 ∗ 0.3 ∗ 0.7 = 0.6
b) 0
注:各种‘return to the origin’的概念和计算方法要掌握。
6. If the correlation coefficient of two Gaussian random variables 𝑋 and 𝑌 is 1/a, then which
is correct?
a) When a=2, X and Y independent.
b) When a=∞, X and Y independent.
c) a can be 0.5。
B
7. It is known that the two Gaussian random variables follows distribution 𝑁(0, 2) and
1
𝑁(0, 18), and their correlation coefficient is , then the covariance of them is:
2
a) 1/72
b) 1/12
c) 18
d) 3
D
大题
1. There are 2 boxes B1 and B2, each box contains 100 balls. The first box (B1) has 50 red
balls and 50 blue balls, the second box (B2) has 40 red balls and 60 blue balls. Suppose a box
is selected from the 2 boxes randomly (with probability 0.5, 0.5), and one ball is picked out
from the box randomly. What is the probability that it is a red ball? If the ball picked out is
red, what is the probability that it comes from box B1?
———————————————————————————————————————
Sol
Let 𝑅1= “red ball is picked out”, then
𝑃(𝑅1|𝐵1) = 0.5, 𝑃(𝑅1|𝐵2 ) = 0.5,
𝑃(𝑅1) = 0.5 ∗ 0.5 + 0.5 ∗ 0.4 = 0.45
𝑃(𝑅1|𝐵1 )𝑃(𝐵1 ) 0.25 5
𝑃(𝐵1|𝑅1) = = =
𝑃(𝑅1) 0.45 9
———————————————————————————————————————
注:条件概率相关知识点要掌握。
𝑥 𝑦 , 0 ≤ 𝑥 ≤ 1, 0 ≤ 𝑦 ≤ 1
2. Given 𝑓(𝑥, 𝑦) = { ,calculate 𝑓𝑋|𝑌 (𝑥|𝑦), 0 ≤ 𝑦 ≤ 1.
0, others
———————————————————————————————————————
Solution:
1 1 1
𝑓𝑌 (𝑦) = ∫0 𝑥 𝑦 𝑑𝑥 = y+1 (𝑥 𝑦+1 )|10 = y+1,0 ≤ 𝑦 ≤ 1
Then, when 0 ≤ 𝑦 ≤ 1:
𝑓(𝑥, 𝑦) (𝑦 + 1)𝑥 𝑦 , 0 ≤ 𝑥 ≤ 1,0 ≤ 𝑦 ≤ 1
𝑓𝑋|𝑌 (𝑥|𝑦) = ={
𝑓𝑌 (𝑦) 0, 𝑜𝑡ℎ𝑒𝑟𝑠
注:
⚫ 可以尝试计算 𝑓𝑌|𝑋 (𝑦|𝑥) =?; 𝑓𝑋|𝑌 (𝑥|𝑦) =?
⚫ 还有例如这类问题:
𝑐𝑥 𝑦 , 0 ≤ 𝑥 ≤ 1, 0 ≤ 𝑦 ≤ 1
Given 𝑓(𝑥, 𝑦) = { , calculate 𝑐?
0, others
⚫ 对二维函数:给𝑓(𝑥, 𝑦),求𝐹(𝑥, 𝑦)要懂得怎么求解;以及𝑃(𝑋 ≤ 𝑥, Y ≤ 𝑦),𝑃(𝑋 ≥ 𝑥, Y ≥
𝑦)代表了什么?
⚫ 对一维函数:给𝑓(𝑥),求𝐹(𝑥)要懂得怎么求解;以及𝑃(𝑋 ≤ 𝑥),𝑃(𝑋 ≥ 𝑥)代表了什么?
𝑋 Marginal Distribution Y
-1 1
𝑌
-1 a 0.3 b
1 0.3 0.2 0.5
Marginal
Distribution 0.5 0.5
X
a) Calculate a and b
b) Calculate the covariance of X and Y
———————————————————————————————————————
Solution:
a) a=0.2, b=0.5
b) and:
𝐸(𝑋) = −1 × 0.5 + 1 × 0.5 = 0,𝐸(𝑌) = −1 × 0.5 + 1 × 0.5 = 0
𝐸(𝑋𝑌) = (−1) × (−1) × 0.2 + (−1) × 1 × 0.3 + 1 × (−1) × 0.3 + 1 × 1 × 0.2 = −0.2
𝑐𝑜𝑣( 𝑋, 𝑌) = 𝐸[(𝑋 − 𝐸(𝑋)(𝑌 − 𝐸(𝑌)] = 𝐸(𝑋𝑌) − 𝐸(𝑋)𝐸(𝑌) = −0.2 − 0 = −0.2
———————————————————————————————————————
5. It is known that the pdf of the random variable X is 𝑓𝑋 (𝑥), and 𝑌 = 𝑋, calculate 𝑓𝑋𝑌 (𝑥, 𝑦)。
———————————————————————————————————————
Solution: to calculate distributions, we can start from definition:
𝐹𝑋𝑌 (𝑥, 𝑦) = 𝑃(𝑋 ≤ 𝑥, 𝑌 ≤ 𝑦) = 𝑃(𝑋 ≤ 𝑥, 𝑋 ≤ 𝑦)
= 𝐹𝑋 (𝑥)𝑈(𝑦 − 𝑥) + 𝐹𝑋 (𝑦)𝑈(𝑥 − 𝑦)
This is because, when 𝑦 ≥ 𝑥, then any 𝑋 ≤ 𝑥 leads to 𝑋 ≤ 𝑦 and thus 𝐹𝑋𝑌 (𝑥, 𝑦) = 𝐹𝑋 (𝑥);
otherwise, 𝐹𝑋𝑌 (𝑥, 𝑦) = 𝐹𝑋 (𝑦). In this case, we should use the Step function in order to take
derivative. Otherwise, it is hard to represent the result of 𝑦 = 𝑥.
Take derivative twice and get
𝑓𝑋𝑌 (𝑥, 𝑦) = 𝑓𝑋 (𝑥)𝛿(𝑦 − 𝑥)
———————————————————————————————————————
6. Let 𝑋(𝑡) = sin (𝜔0 𝑡 + Φ), 𝑌(𝑡) = cos (𝜔0 𝑡 + Φ), where 𝜔0 is a constant, Φ uniformly
distributed in (−π,π), calculate the cross-covariance 𝐶𝑋𝑌 。
———————————————————————————————————————
Solution:
1 π
𝐸(𝑋(𝑡)) = 𝐸(𝑠𝑖𝑛(𝜔0 𝑡 + Φ)) = ∫ 𝑠𝑖𝑛(𝜔0 𝑡 + φ) dφ = 0
2π −π
Similarly:
𝐸(𝑌(𝑡)) = 0
And
1 1
𝐶𝑋𝑌 = 𝑅𝑋𝑌 − 𝑚𝑋 𝑚𝑌 = 𝐸{𝑠𝑖𝑛(𝜔0 𝑡1 + 𝜔0 𝑡2 + 2𝛷) + 𝑠𝑖𝑛(𝜔0 (𝑡1 − 𝑡2 ))} = 𝑠𝑖𝑛(𝜔0 𝜏)
2 2
𝜏 = 𝑡1 − 𝑡2
———————————————————————————————————————
1
7. The autocovariance of the process 𝑋(𝑡) is 𝐶𝑋 (𝜏) = 4 𝑒 −2𝜆|𝜏| , please
———————————————————————————————————————
Solution:
1 𝐶𝑋 (𝜏)
1) 𝜎𝑋2 = 𝐶𝑋 (0) = 4, 𝑟𝑋 (𝜏) = 2
𝜎𝑋
= 𝑒 −2𝜆|𝜏| ,
∞ ∞ 1
Therefore 𝜏0𝑋 = ∫0 𝑟𝑋 (𝜏) 𝑑𝜏 = ∫0 𝑒 −2𝜆𝜏 𝑑𝜏 = 2𝜆
1 𝐶𝑋 (𝜏)
2) 𝐶𝑋 (𝜏) = 𝑒 −2𝜆|𝜏| , 𝑟𝑋 (𝜏) = 2 = 𝑒 −2𝜆|𝜏| ,
4 𝜎𝑋
𝜋
𝜋
Therefore 𝑟𝑋 ( ) = 𝑒 −2𝜆|𝜆| = 𝑒 −2𝜋
𝜆
———————————————————————————————————————
8. The autocorrelation of the stationary stochastic process 𝑋(𝑡) is 𝑅𝑋 (𝜏) = 100𝑒 −10|𝜏| + 100,
find 𝐸{𝑋 (𝑡)2} and 𝑉𝑎𝑟{𝑋 (𝑡)}.
———————————————————————————————————————
Solution:
The square of the mean is 𝑚𝑋22 = 𝑅𝑋2 (∞) = 100
mean square 𝐸[𝑋 2 (𝑡)] = 𝑅𝑋 (0) = 200
Variance 𝑉𝑎𝑟{𝑋 (𝑡)} = 𝜎𝑋2 = 𝑅𝑋 (0) − 𝑚𝑋2 = 100
———————————————————————————————————————
9. The real stationary discrete-time process 𝒙[𝑛] = [1,2,3], calculate 𝑅𝒙 [𝑚] and 𝑆𝑥 (𝜔)
———————————————————————————————————————
Solution:
𝑅𝒙 [𝑚] = [3,4,14/3,4, 3] for 𝑚 = −2, −1,0,1,2, and 0 for others.
1∗1+2∗2+3∗3
𝑅𝒙 [0] = = 14/3
3
2∗1+3∗2
𝑅𝒙 [1] = =4
2
∞ ∞
−𝑗𝑚𝜔
𝑆(𝜔) = ∑ 𝑅[𝑚]𝑒 = 𝑅[0] + 2 ∑ 𝑅[𝑚] cos(𝜔𝑚) = 14/3 + 8 cos(𝜔) + 6 cos(2𝜔)
𝑚=−∞ 𝑚=1
———————————————————————————————————————
注:
⚫ 给𝑥[𝑛]求𝑅𝑥 [𝑚];给𝑅𝑥 [𝑚]和𝑆𝑥 (𝜔)之一求解另外一个,都是离散随机过程的基础知识点,
一定要掌握。
⚫ 𝑆(𝜔) = ∑∞𝑚=−∞ 𝑅[𝑚]𝑒
−𝑗𝑚𝜔
= 𝑅[0] + 2 ∑∞
𝑚=1 𝑅[𝑚] cos(𝜔𝑚),其实是两条公式,都要掌
握。同理,逆傅里叶变换也是有两条公式,也要掌握。
⚫ 此外,例如,给𝑆𝑥 (𝜔)求𝑅𝑥 [1],给𝑅𝑥 [𝑚]求𝑆𝑥 (𝜋)这种求𝑚和𝜔在一个特定值下的过程也
要掌握。
⚫ 特别地,复数的情况也要注意。例如,𝑥[𝑛] = [1 + j, 2 − j, 3 + j], calculate 𝑅𝑥 [𝑚] and
𝑆𝑥 (𝜔)?
10. Given joint stationary processes 𝑋(𝑡) and 𝑌(𝑡) . 𝑌(𝑡) = 𝑋(𝑡) + 2𝑋(𝑡 − 1) . the
autocorrelation of 𝑋(𝑡) is 𝑅𝑋 (𝜏) = 𝑒 −|𝜏| . Calculate 𝑅𝑌𝑋 (𝜏).
———————————————————————————————————————
Solution:
𝑅𝑌𝑋 (𝜏) = 𝐸{[𝑋(𝑡 + 𝜏) + 2𝑋(𝑡 − 1 + 𝜏)]𝑋(𝑡)} = 𝑅𝑋 (𝜏) + 2𝑅𝑋 (𝜏 − 1)
= 𝑒 −|𝜏| + 2𝑒 −|𝜏−1|
———————————————————————————————————————
注:
⚫ 进一步地,𝑅𝑌 (𝜏) =?
⚫ 还可以进一步地引申一些复杂的情况,例如𝑌1 (𝑡) = 𝑋(𝑡) + 2𝑋(𝑡 − 1),𝑌2 (𝑡) = 𝑋(𝑡) +
2𝑋(𝑡 − 2),𝑌3 (𝑡) = 𝑋(𝑡) + 2𝑋(𝑡 − 3),然后求𝑅𝑌1 𝑌2 (𝜏),𝑅𝑌1 𝑌3 (𝜏),𝑅𝑌2 𝑌3 (𝜏),𝑅𝑌3 (𝜏)等等。
⚫ 又例如,给出𝑅𝑌1 𝑌2 (𝜏),𝑅𝑌1 𝑌3 (𝜏),𝑅𝑌2 𝑌3 (𝜏),𝑅𝑌3 (𝜏)等等,然后已知𝑌1 (𝑡) = 𝑋(𝑡) + 2𝑋(𝑡 − 1),
求𝑌2 (𝑡) =? ?这种。
⚫ 最后,例如不给𝑅𝑋 (𝜏)的表达式,而是把𝑅𝑋 (𝜏)当做一个已知变量,求𝑅𝑌1 𝑌2 (𝜏),𝑅𝑌1 𝑌3 (𝜏),
𝑅𝑌2 𝑌3 (𝜏),𝑅𝑌3 (𝜏)等等(以𝑅𝑋 (𝜏)表示),甚至,让𝑅𝑌1 𝑌2 (𝜏)当做一个已知变量,求𝑅𝑌1 𝑌3 (𝜏)
(以𝑅𝑌1 𝑌2 (𝜏)表示) 。都是很有趣的题目。
11. A white noise with power spectrum 𝑞 passes through an RC system as follows (input 𝑥(𝑡),
output 𝑦(𝑡)):
𝑑𝑦(𝑡)
𝑅𝐶 + 𝑦(𝑡) = 𝑥(𝑡)
𝑑𝑡
Where R>0 and C>0.
a) Calculate the impulse response ℎ(𝑡)。
b) Calculate the mean 𝑚𝑌 , autocorrelation 𝑅𝑌 (𝜏) and power spectrum 𝑆𝑌 (𝜔) of the output
c) Calculate 𝑅𝑌𝑋 (𝜏) and 𝑆𝑌𝑋 (𝜔)。
———————————————————————————————————————
Solution:
a)
1
1 𝑅𝐶
𝐻(𝜔) = = 1
𝑅𝐶𝑗𝜔 + 1 + 𝑗𝜔
𝑅𝐶
1 −𝑡
ℎ(𝑡) = 𝑒 𝑅𝐶 𝑈(𝑡)
𝑅𝐶
1 −𝑡
ℎ(𝑡) = {𝑅𝐶 𝑒 , 𝑡 ≥ 0
𝑅𝐶
0 ,𝑡 < 0
∞ 1
b) 𝑚𝑌 = 𝑚𝑥 ∫0 ℎ(𝑡) 𝑑𝑡 = 0, |𝐻(𝜔)|2 = 𝑅2 𝐶 2𝜔2+1
𝑞
→ 𝑆𝑌 (𝜔) = |𝐻(𝜔)|2 𝑆𝑋 (𝜔) =
𝑅2 𝐶 2 𝜔2 +1
1
𝑞
→ 𝑅𝑌 (𝜏) = 2𝑅𝐶 𝑒 −𝑅𝐶|𝜏|
c)
𝑞 𝑞 1
𝑆𝑌𝑋 (𝜔) = 𝑆𝑋 (𝜔)𝐻(𝜔) = = 1
1 + 𝑗𝑅𝐶𝜔 𝑅𝐶 + 𝑗𝜔
𝑅𝐶
𝜏
𝑞
𝑅𝑌𝑋 (𝜔) = 𝑅𝐶 𝑒 −𝑅𝐶 𝑈(𝜏)
Note:
q q 1
𝑆𝑋𝑌 (𝜔) = 𝑆𝑋 (𝜔)𝐻 ∗ (𝜔) = = 1
1 − 𝑗𝑅𝐶𝜔 𝑅𝐶 − 𝑗𝜔
𝑅𝐶
𝑞 𝜏
𝑅𝑋𝑌 (𝜏) = 𝑒 𝑅𝐶 𝑈(−𝜏)
𝑅𝐶
———————————————————————————————————————
𝜔2 +10
12. Given power spectrum 𝑆𝑋 (𝜔) = , calculate the autocorrelation, mean, average
𝜔4 +20𝜔2 +64
1 1
(𝑤 2 + 4) + (𝑤 2 + 16) 1 1 1 1 1 2×2 1 2×4
2 2
S𝑋 (𝜔) = = + = +
(𝑤 2 + 4)(𝑤 2 + 16) 2 2 2
2 𝜔 + 4 2 𝜔 + 16 8 𝜔 + 2 2 16 𝜔 2 + 42
→
1 1 2α
𝑅𝑋 (𝜏) = 𝑒 −2|𝜏| + 𝑒 −4|𝜏| (e−α|τ| ↔ )
8 16 w2 +α2
3 3
Variance: σ2X = 𝑅𝑋 (0) − 𝑚𝑋2 = −0 =
16 16
1 −2|𝜏| 1 −4|𝜏|
2 𝑒 + 𝑒
𝐶𝑋 (𝜏) 𝑅𝑋 (𝜏)−𝑚𝑋 𝑅𝑋 (𝜏) 2 1
Correlation coefficient: 𝑟(𝜏) = σ2X
= σ2X
= σ2X
= 8
3
16
= 3 𝑒 −2|𝜏| + 3 𝑒 −4|𝜏|
16
∞ ∞ 2 1 2 1 1 1 5
correlation time: 𝜏0 = ∫0 𝑟(𝜏)𝑑𝜏 = ∫0 ( 𝑒 −2𝜏 + 𝑒 −4𝜏 ) 𝑑𝜏 = ⋅ + ⋅ =
3 3 3 2 4 3 12
———————————————————————————————————————
注:
𝑑𝜔2 +𝑒 𝑔𝜔4 +𝑑𝜔2 +𝑒
⚫ 作为练习,可以尝试各种通用的𝑆𝑋 (𝜔) = 𝑎𝜔4 +𝑏𝜔2 +𝑐; 𝑆𝑋 (𝜔) = 𝑓𝜔6 +𝑎𝜔4 +𝑏𝜔2 +𝑐这些情况。
𝜔2 +10
⚫ 除了𝑆𝑋 (𝜔) = 𝜔4 +20𝜔2 +64这种二项式的类型,还有例如𝑆𝑋 (𝜔) = 1(常数),冲击函数,
各种三角函数,这些也要熟悉:
13. Let the stochastic process 𝑋(𝑡) = 𝐴𝑡, where 𝐴 is a random variable with standard normal
distribution. Is 𝑋(𝑡) Wide-Sense Stationary? Prove it.
———————————————————————————————————————
Solution:
𝐸{𝑋(𝑡)} = 𝐸{𝑡𝐴} = 𝑡𝐸{𝐴} = 0
𝑅𝑋 (𝑡1 , 𝑡2 ) = 𝐸{𝑋(𝑡1 )𝑋(𝑡2 )} = 𝑡1 𝑡2𝐸{𝐴2 } = 𝑡1 𝑡2
Therefore 𝑋(𝑡) is not WSS.
———————————————————————————————————————
注:证明一个过程是否 WSS,或者是否 ergodicity process,是随机过程里面的一个基本问题。
注意,一般地,证明是否 WSS 看的是 mean 和 autocorrelation,求解出来即可;而如果出现
需要证明是否 SSS,除非从分布证明,否则一般是举例。
4λ
14. Given the input Gaussian stationary process 𝑋(𝑡) with power spectrum SX (ω) = 4λ2 +ω2,
input 𝑋(𝑡) to the system below, calculate the 1-D pdf of the output 𝑌(𝑡)
a 2b a
Hint: ∫−a b2 +ω2 dω = 4arctan (b)
———————————————————————————————————————
Solution:
4λ
As SX (ω) = 4λ2 +ω2, we have
𝑅𝑋 (𝜏) = 𝑒 −2𝜆|𝜏|
Input Gaussian, and the system is a deterministic system, then the output is also a Gaussian
process.
2 (2𝜆)
𝑆𝑌 (𝜔) = 𝑆𝑋 (𝜔)|𝐻(𝜔)|2 =
(2𝜆)2 + 𝜔 2
1 𝛥𝜔 2 (2𝜆) 2 𝛥𝜔
𝑅𝑌 (0) = ∫ 𝑑𝜔 = 𝑎𝑟𝑐𝑡𝑎𝑛 ( )
2𝜋 −𝛥𝜔 (2𝜆)2 + 𝜔 2 𝜋 2𝜆
𝑚𝑋 2 = 𝑅𝑋 (+∞) = 0, 𝑚𝑋 = 0
+∞
𝑚𝑌 = 𝑚𝑋 ∫ ℎ(𝜏) 𝑑𝜏 = 0
0
2 Δω
𝜎𝑌 2 = 𝑅𝑌 (0) − 𝑚𝑌 2 = 𝑎𝑟𝑐𝑡𝑎𝑛( )
𝜋 2λ
Therefore the 1-D pdf of the output 𝑌(𝑡) is:
πy 2
𝑦2 −
1 −
2𝜎𝑌 2
1 Δω
4 𝑎𝑟𝑐𝑡𝑎𝑛( 2λ )
𝑃(𝑦) = 𝑒 = 𝑒
√2𝜋𝜎𝑌 √4 𝑎𝑟𝑐𝑡𝑎𝑛(
Δω
)
2λ
———————————————————————————————————————
注:
⚫ 在本题里面,是用画图来表示 power spectrum,那么,如果换成表达式?是否还能求解?
⚫ 此外,如果是先给𝑅𝑋 (𝜏),求了𝑆𝑋 (𝜔),再求 1-D pdf of the output,也要懂。这是一系列
知识点,而不是一个个的零散的知识点。
𝑒 −𝑎𝑡 , 𝑡 ≥ 0,
b) ℎ(𝑡) = {
0, 𝑡<0
c) 𝑌(𝑡) = 𝑋(𝑡) + 2𝑋(𝑡 − 𝜏)
———————————————————————————————————————
Solution:
(a)
3𝑗𝜔 + 4 𝑗3𝜔 + 4
𝐻(𝜔) = 2
= 2
(jw) − 3𝑗𝜔 + 2 −𝜔 − 𝑗3𝜔 + 2
9𝜔2 + 16 9𝜔2 + 16
|𝐻(𝜔)|2 = =
(2 − 𝜔 2 )2 + 9𝜔 2 𝜔 4 + 5𝜔 2 + 4
(b)
+∞ +∞
1
𝐻(𝜔) = ∫ ℎ(𝑡)𝑒 −𝑗ωt 𝑑𝑡 = ∫ 𝑒 −𝑎𝑡 𝑒 −𝑗ωt 𝑑𝑡 =
−∞ 0 𝑎 + 𝑗𝜔
1
|𝐻(𝜔)|2 = 2
𝑎 + 𝜔2
(c)
𝑌(𝜔) = 𝑋(𝜔) + 2𝑋(𝜔)𝑒 −𝑗ωτ
𝑌(𝜔) 1 + 2𝑒 −𝑗ωτ
𝐻(𝜔) = = = 1 + 2𝑒 −𝑗ωτ
𝑋(𝜔) 1
|𝐻(𝜔)|2 = (1 + 2𝑒 −𝑗𝜔𝜏 )(1 + 2𝑒 𝑗𝜔𝜏 ) = 1 + 4 + 2(𝑒 −𝑗𝜔𝜏 + 𝑒 𝑗𝜔𝜏 ) = 5 + 4cos (𝜔𝜏)
———————————————————————————————————————
𝑑𝑌(𝑡)
16. For the joint stationary processes X(t) and Y(t): + 𝑌(𝑡) = 𝑋(𝑡), and the power
𝑑𝑡
1
spectrum of X(t) is SX (ω)= ω2+1 , calculate the SYX (ω), SXY (ω) and SY (ω)
———————————————————————————————————————
Solution:
𝑗𝜔𝑌(𝜔) + 𝑌(𝜔) = 𝑋(𝜔),
1
→𝐻𝑌 (𝜔) = 𝑗𝜔+1
1 1
→𝑆𝑌𝑋 (𝜔) = 𝐻𝑌 (𝜔)𝑆𝑋 (𝜔) = 𝑗𝜔+1 𝜔2 +1 。
1 1
→𝑆𝑋𝑌 (𝜔) = 𝐻𝑌∗ (𝜔)𝑆𝑋 (𝜔) = −𝑗𝜔+1 𝜔2 +1 。
1 2
→𝑆𝑌 (𝜔) = |𝐻𝑌 (𝜔)|2 𝑆𝑋 (𝜔) = (𝜔2 +1 )
———————————————————————————————————————
注:
𝑑𝑌1 (𝑡) 𝑑𝑌2 (𝑡)
⚫ 可以进一步地引申一些复杂的情况,例如 + 𝑌1 (𝑡) = 𝑋(𝑡), − 𝑌2 (𝑡) = 𝑋(𝑡),
𝑑𝑡 𝑑𝑡
𝑑𝑌3 (𝑡)
+ 𝑌3 (𝑡) = 𝑌1 (𝑡)等等,然后求𝑆𝑌1 𝑌2 (𝜔),𝑆𝑌3 (𝜔)等等。又例如:
𝑑𝑡
𝑑𝑌2 (𝑡)
− 𝑌2 (𝑡) = 𝛽𝑋(𝑡),求𝛼, 𝛽,也很常见
𝑑𝑡
⚫ 这些都是很有趣的题目
17. A white noise with power spectrum N0 /2 is inputted to a band pass filter H(ω) and get the
output Y(t), whose physical spectrum FY (ω) is shown as the following figure. Calculate
the noise equivalent pass-band Δfe (value only, unit omitted).
———————————————————————————————————————
Solution:
a) 𝐹𝑌 (𝜔) is:
𝑁0 [𝜔 − (𝜔0 − 2)], (𝜔0 − 2 ≤ 𝜔 ≤ 𝜔0 )
𝐹𝑌 (𝜔) = { 𝑁0 [(𝜔0 + 2) − 𝜔], (𝜔0 ≤ 𝜔 ≤ 𝜔0 + 2)
0, (𝑜𝑡ℎ𝑒𝑟𝑠)
or
𝑁0 [2 − |𝜔 − 𝜔0 |], (−2 ≤ 𝜔 − 𝜔0 ≤ 2)
𝐹𝑌 (𝜔) = { ⬚
0, (𝑜𝑡ℎ𝑒𝑟𝑠)
Thus
∞
1 ∫0 𝐹𝑌 (𝜔)𝑑𝜔 2 𝜔0 1
𝛥𝑓𝑒 = = ∫ [𝜔 − (𝜔0 − 2)]𝑑𝜔 =
2𝜋 𝐹𝑌 (𝜔0 ) 4𝜋 𝜔0 −2 𝜋
———————————————————————————————————————
2
18. A Gaussian white noise with power spectrum 𝑞 is inputted to a filter 𝐻(𝜔) = , and get
2+𝑗𝜔
√2𝜋𝜎𝑌2 √2𝜋𝑞
———————————————————————————————————————
19. Given 𝑌(𝑡) = 𝑋(𝑡) + 𝑁(𝑡), where 𝑋(𝑡) is a deterministic signal with spectrum 𝑋(𝜔) as
follows. 𝑁(𝑡) is a Gaussian white noise with power spectrum 𝑞.
𝑗𝜔, (0 ≤ 𝜔 ≤ 1)
𝑋(𝜔) = {
0, (𝑜𝑡ℎ𝑒𝑟𝑠)
let 𝑌(𝑡) passes through its matched filter 𝐻(𝜔), and calculate the maximum signal-to-noise
ratio of the output.
———————————————————————————————————————
Solution:
We have 𝑆𝑁 (𝜔) = 𝑞
𝜔2 , (0 ≤ 𝜔 ≤ 1)
|𝑋(𝜔)|2 = {
0, (𝑜𝑡ℎ𝑒𝑟𝑠)
+∞ 1
1 1
∫ |𝑋(𝜔)|2 𝑑𝜔 = ∫ 𝜔2 𝑑𝜔 = (1 − 0) =
−∞ 0 3 3
Then the maximum signal-to-noise ratio is:
1 +∞ |𝑋(𝜔)|2 1 1 1
𝑑𝑚 = ∫ 𝑑𝜔 = =
2𝜋 −∞ 𝑆𝑁 (𝜔) 2𝜋𝑞 3 6𝜋𝑞
———————————————————————————————————————
1
注:求各种比例和各种时延条件下的 maximum signal-to-noise ratio:𝑌(𝑡) = 2 𝑋(𝑡) + 𝑁(𝑡)?
20. The autocorrelation function 𝑋(𝑡) of a zero-mean stationary normal stochastic process is
𝑅𝑋 (𝜏) = e−2|𝜏| , determine its ergodicity.
———————————————————————————————————————
Solution:
As it is zero mean white Gaussian, perform
+∞ +∞
1
∫ 𝑅𝑋 (𝜏) 𝑑𝜏 = ∫ e−2𝜏 𝑑𝜏 = < ∞
0 0 2
And thus it is an ergodicity process
———————————————————————————————————————
注:
⚫ 证明一个过程是否 WSS,或者是否 ergodicity process,是随机过程里面的一个基本问题。
⚫ 一般地,证明是否 WSS 看的是 mean 和 autocorrelation,求解出来即可;而如果出现需
要证明是否 SSS,除非从分布除非,否则,一般是举例。
⚫ 而 ergodicity process 的前提就是 WSS。
⚫ 对 ergodicity 的证明有两种方法:一是使用上述的判断条件,仅在 zero-mean stationary
normal/Gaussian stochastic process 的前提下成立;二是看统计平均和时间平均,统计自
相关和时间自相关是否相等。
⚫ 例 如 , 𝑌(𝑡) = 𝑐𝑜𝑠(𝜔0 𝑡 + 𝜃), where 𝜔0 is a constant, 𝜃 is uniformly distributed in
(−𝜋, 𝜋)。对𝑌(𝑡),最终可以证明其是 ergodicity process。
但是,如果𝜃 is uniformly distributed in (0, 1/2 𝜋),因为其在求“时间自相关”时,相
位𝜃所经历的状态不是周期的倍数,导致其无法收敛到τ,和 “统计自相关”不同,因
此不是 ergodicity process。实际上,它甚至不是 WSS 的。此外,大家也可以如作业 3
中,𝑌(𝑡) = 𝑋𝑐𝑜𝑠(𝜔0 𝑡 + 𝜃)的各种情况
进一步的,如果题目说 𝜃 is uniformly distributed in (𝑎, 𝑏),然后是 WSS 和 ergodicity
process,那么𝑎, 𝑏要符合什么条件?也很值得思考。
更进一步地,如果不是实数,是复数又会如何?例如𝑌(𝑡) = 𝑋𝑒 𝑗(𝜔0 𝑡+𝜃) ?
21. A plays a game consecutively with B as follows: in each time (of playing the game), the
probabilities of A win and lose are 0.4 and 0.6. A starts with 1 point, B starts with 2 points, and
in each time (of playing the game), the winner record ‘+1’ and the loser ‘-1’ point. The game
will stop when A lost all his/her points (points equals to 0, defined as A losing the whole game)
or reach 3 points (points equals to 3, defined as A winning the whole game). What’s the
probability of A losing the whole game within 2 times (lose after 1 or 2 games)?
———————————————————————————————————————
Solution:
1 0 0 0
The one step transition matrix for A is 𝑃 = [ 0.6 0 0.4 0]
0 0.4 0 0.6
0 0 0 1
1 0 0 0 1 0 0 0 1 0 0 0
2
𝑃 =[ 0.6 0 0.4 0 ][0.6 0 0.4 0 ]=[ 0.6 0.24 0 0.16 ]
0 0.6 0 0.4 0 0.6 0 0.4 0.36 0 0.24 0.4
0 0 0 1 0 0 0 1 0 0 0 1
the probability of A losing the whole game within 2 times is 0.6
———————————————————————————————————————
注:马尔可夫链中,各种计算所代表的概念要理解。