ECS 452: Digital Communication Systems 2015/2. HW 1 Due: Feb 5

Size: px

Start display at page:

Download "ECS 452: Digital Communication Systems 2015/2. HW 1 Due: Feb 5"

Wilfred Lawson
5 years ago
Views:

ECS 452: Digital Communication Systems 2015/2 HW 1 Due: Feb 5 Lecturer: Asst. Prof. Dr. Prapun Suksompong Instructions (a) Must solve all non-optional problems.

1 ECS 452: Digital Communication Systems 2015/2 HW 1 Due: Feb 5 Lecturer: Asst. Prof. Dr. Prapun Suksompong Instructions (a) Must solve all non-optional problems. (5 pt) (i) Write your first name and the last three digit of your student ID on the upper-right corner of every submitted page. (ii) For each part, write your explanation/derivation and answer in the space provided. (b) ONE part of a question will be graded (5 pt). Of course, you do not know which part will be selected; so you should work on all of them. (c) Late submission will be rejected. (d) Write down all the steps that you have done to obtain your answers. You may not get full credit even when your answer is correct without showing how you get your answer. Problem 1 (HW ). Consider the code {0, 01} (a) Is it nonsingular? (b) Is it uniquely decodable? (c) Is it prefix-free? 1-1

2 ECS 452 HW 1 Due: Feb /2 Problem 2 (HW ). Consider the random variable X whose support S X seven values: S X = {x 1, x 2,..., x 7 }. Their corresponding probabilities are given by contains (a) Find the entropy H(X). x x 1 x 2 x 3 x 4 x 5 x 6 x 7 p X (x) (b) Find a binary Huffman code for X. (c) Find the expected codelength for the encoding in part (b). 1-2

ECS 452 HW 1 Due: Feb 5 2015/2 Problem 3

3 ECS 452 HW 1 Due: Feb /2 Problem 3 (HW ). Find the entropy and the binary Huffman code for the random variable X with pmf { x, x = 1, 2,..., 6, p X (x) = 21 0, otherwise. Also calculate E [l(x)] when Huffman code is used. Problem 4 (HW ). These codes cannot be Huffman codes. Why? (a) {00, 01, 10, 110} (b) {01, 10} (c) {0, 01} 1-3

4 ECS 452 HW 1 Due: Feb /2 Problem 5 (HW ). A memoryless source emits two possible message Y(es) and N(o) with probability 0.9 and 0.1, respectively. (a) Determine the entropy (per source symbol) of this source. (b) Find the expected codeword length per symbol of the Huffman binary code for the third-order extensions of this source. (c) Use MATLAB to find the expected codeword length per (source) symbol of the Huffman binary code for the fourth-order extensions of this source. (i) Put your answer here. (ii) Don t forget to attach the printout of your MALTAB script (highlighting the modified parts if you start from the provided class example) and the expressions/results displayed in the command window. (d) Use MATLAB to plot the expected codeword length per (source) symbol of the Huffman binary code for the nth-order extensions of this source for n = 1, 2,..., 8. Attach the printout of your plot. 1-4

6 ECS 452 HW 1 Due: Feb /2 Problem 6 (HW ). (Optional) The following claim is sometimes found in the literature: It can be shown that the length l(x) of the Huffman code of a symbol x with probability p X (x) is always less than or equal to log 2 p X (x). Even though it is correct in many cases, this claim is not true in general. Find an example where the length l(x) of the Huffman code of a symbol x is greater than log 2 p X (x). Hint: Consider a pmf that has the following four probability values {0.01, 0.30, 0.34, 0.35}. Problem 7 (HW ). (Optional) Construct a random variable X (by specifying its pmf) whose corresponding Huffman code is {0, 10, 11}. 1-5

ECS 452: Digital Communication Systems 2015/2 HW 2 Due: Feb 19 Lecturer: Asst. Prof. Dr. Prapun Suksompong Instructions (a) Must solve all non-optional problems.

7 ECS 452: Digital Communication Systems 2015/2 HW 2 Due: Feb 19 Lecturer: Asst. Prof. Dr. Prapun Suksompong Instructions (a) Must solve all non-optional problems. (5 pt) (i) Write your first name and the last three digit of your student ID on the upper-right corner of every submitted page. (ii) For each part, write your explanation/derivation and answer in the space provided. (b) ONE part of a question will be graded (5 pt). Of course, you do not know which part will be selected; so you should work on all of them. (c) Late submission will be rejected. (d) Write down all the steps that you have done to obtain your answers. You may not get full credit even when your answer is correct without showing how you get your answer. Problem 1. In this question, each output string from a DMS is encoded by the following source code: (a) Is the code prefix-free? x Codeword c(x) a 1 d 01 e 0000 i 001 o u (b) Is the code uniquely decodable? 2-1

ECS 452 HW 2 Due: Feb 19 2015/2 (c) Suppose the DMS produces the string audio. Find the output of the source encoder.

Use / to indicate the locations where the sting above is split into codewords. Problem 2. A DMC has X = {0, 1} and Y = {1, 2, 3}. The following decoding table is used to decode the channel output.

8 ECS 452 HW 2 Due: Feb /2 (c) Suppose the DMS produces the string audio. Find the output of the source encoder. (d) Suppose the output of the source encoder is Find the corresponding source string produced by the DMS. Use / to indicate the locations where the sting above is split into codewords. Problem 2. A DMC has X = {0, 1} and Y = {1, 2, 3}. The following decoding table is used to decode the channel output. y ˆx(y) Suppose the channel output string is (a) Find the corresponding decoded string. (b) Suppose the channel input string is produced from an ASCII source encoder by the command dec2bin(sourcestring,7) in MATLAB. Assume that there is no channel decoding error. Find the corresponding source string. Problem 3 (HW ). Consider a BSC whose crossover probability for each bit is p = Suppose P [X = 0] = (a) Draw the channel diagram. 2-2

ECS 452 HW 2 Due: Feb 19 2015/2 (b) Find the

reasonable detectors for this binary channel.

P ˆX = 0 X = 0 [ ] P ˆX = 1 X = 1 (f) Find the

9 ECS 452 HW 2 Due: Feb /2 (b) Find the channel matrix Q. (c) Find the joint pmf matrix P. (d) Find the row vector q which contains the pmf of the channel output Y. (e) We now analyze the performance of all four reasonable detectors for this binary channel. Complete the table below: ˆx(y) y 1 y 1 0 [ ] P ˆX = 0 X = 0 [ ] P ˆX = 1 X = 1 (f) Find the MAP detector and its error probability. P (C) P (E) (g) Find the ML detector and its error probability. 2-3

ECS 452 HW 2 Due: Feb 19 2015/2 Problem 4

Consider a DMC whose X = {1, 2, 3}, Y = {1,

Suppose the input probability vector is p =

10 ECS 452 HW 2 Due: Feb /2 Problem 4 (HW ). Consider a DMC whose X = {1, 2, 3}, Y = {1, 2, 3}, and Q = Suppose the input probability vector is p = [0.2, 0.4, 0.4] (a) Find the joint pmf matrix P. (b) Find the row vector q which contains the pmf of the channel output Y. (c) Find the error probability of the naive decoder. (d) Find the error probability of the (DIY) decoder ˆx(y) = 4 y. 2-4

11 ECS 452 HW 2 Due: Feb /2 (e) Find the MAP detector and its error probability. (f) Find the ML detector and its error probability. Problem 5 (HW ). Consider a BAC whose Q(1 0) = 0.35 and Q(0 1) = Suppose P [X = 0] = 0.4. (a) Draw the channel diagram. (b) Find the joint pmf matrix P. 2-5

ECS 452 HW 2 Due: Feb 19 2015/2 (c) Find the row vector q which contains the pmf of the channel output Y.

12 ECS 452 HW 2 Due: Feb /2 (c) Find the row vector q which contains the pmf of the channel output Y. (d) We now analyze the performance of all four reasonable detectors for this binary channel. Complete the table below: ˆx(y) y 1 y 1 0 [ ] P ˆX = 0 X = 0 [ ] P ˆX = 1 X = 1 P (C) P (E) 2-6

ECS 452 HW 2 Due: Feb 19 2015/2 (e) Find the MAP detector and its error probability.

(Optional) Consider a DMC whose samples of input X and output Y are recorded as row vectors x and y in the file HW DMC Channel

Note that most of these can be solved by appropriate parts of the m-files posted on the course web site. (a) The support X of X.

13 ECS 452 HW 2 Due: Feb /2 (e) Find the MAP detector and its error probability. (f) Find the ML detector and its error probability. Problem 6 (HW ). (Optional) Consider a DMC whose samples of input X and output Y are recorded as row vectors x and y in the file HW DMC Channel Data.mat. Write MATLAB script which uses the recorded information to estimate the quantities below. Note that most of these can be solved by appropriate parts of the m-files posted on the course web site. (a) The support X of X. (b) The support Y of Y. (c) The row vector p which contains the pmf of X. (d) The Q matrix. (e) The row vector q which contains the pmf of Y. Do this using two methods: (i) Count directly from the observed values of Y. (ii) Use the estimated values of p and Q. 2-7

14 ECS 452 HW 2 Due: Feb /2 (f) The error probability when the naive decoder is used. Do this using two methods: (i) Directly construct ˆx from y. Then, compare ˆx and x. (ii) Use the estimated values of p and Q. (g) The error probability when the MAP decoder is used. Do this using two methods: (i) First find the MAP decoder table using the estimated values of p and Q. Then, construct ˆx from y according to the decoder table. Finally, compare ˆx and x. (ii) Use the estimated values of p and Q to directly calculate the error probability. (h) The error probability when the ML decoder is used. Do this using two methods: (i) First find the ML decoder table using the estimated value of Q. Then, construct ˆx from y according to the decoder table. Finally, compare ˆx and x. (ii) Use the estimated values of p and Q to directly calculate the error probability. 2-8

15 ECS452_HW_2015_2_Sol Page 1 Q6 MATLAB Simulation of the system in Q4 Saturday, September 13, 2014 % MATLAB Script for Q6 of HW2 for ECS 452 % By Asst. Prof. Dr. Prapun Suksompong. close all; clear all; tic 11:44 PM load HW_DMC_Channel_Data (a) S_X = unique(x) (b) S_Y = unique(y) %% Statistical Analysis n = length(x); (c) % The probability values for the channel inputs p_x_sim = hist(x,s_x)/n % Relative frequencies from the simulation (d) % The channel transition probabilities from the simulation Q_sim = []; for k = 1:length(S_X) I = find(x==s_x(k)); LI = length(i); rel_freq_xk = LI/n; yc = y(i); cond_rel_freq = hist(yc,s_y)/li; Q_sim = [Q_sim; cond_rel_freq]; end Q_sim % Relative frequencies from the simulation (e) (e.i) p_y_sim = hist(y,s_y)/n % Relative frequencies from the simulation (e.ii) p_y_sim2 = p_x_sim*q_sim %% p_x = p_x_sim; Q = Q_sim; (f) %% Naive Decoder x_hat = y;

16 (f.i) % Error Probability PE_sim_Naive = 1-sum(x==x_hat)/n % Error probability from the simulation (f.ii) % Calculation of the theoretical error probability PC = 0; for k = 1:length(S_X) t = S_X(k); i = find(s_y == t); if length(i) == 1 PC = PC+ p_x(k)*q(k,i); end end PE_theoretical_Naive = 1-PC (g) %% MAP Decoder P = diag(p_x)*q; % Weight the channel transition probability by the % corresponding prior probability. [V I] = max(p); % For I, the default MATLAB behavior is %that when there are multiple max, the % index of the first one is returned. Decoder_Table_MAP = S_X(I) % The decoded values corresponding to the % received Y (g.i) % Decode according to the decoder table x_hat = y; % preallocation for k = 1:length(S_Y) I = (y==s_y(k)); x_hat(i) = Decoder_Table_MAP(k); end PE_sim_MAP = 1-sum(x==x_hat)/n % Error probability from the simulation (g.ii) % Calculation of the theoretical error probability Decoder_Table = Decoder_Table_MAP; PC = 0; for k = 1:length(S_X) I = (Decoder_Table == S_X(k)); Q_row = Q(k,:); PC = PC+ p_x(k)*sum(q_row(i)); end PE_theoretical_MAP = 1-PC (h) %% ML Decoder [V I] = max(q); % For I, the default MATLAB behavior is %that when there are multiple max, the % index of the first one is returned. Decoder_Table_ML = S_X(I) % The decoded values corresponding to the ECS452_HW_2015_2_Sol Page 2

17 ECS452_HW_2015_2_Sol Page 3 % received Y % Decode according to the decoder table x_hat = y; % preallocation for k = 1:length(S_Y) I = (y==s_y(k)); x_hat(i) = Decoder_Table_ML(k); end (h.i) PE_sim_ML = 1-sum(x==x_hat)/n % Error probability from the simulation (h.ii) % Calculation of the theoretical error probability Decoder_Table = Decoder_Table_ML; PC = 0; for k = 1:length(S_X) I = (Decoder_Table == S_X(k)); Q_row = Q(k,:); PC = PC+ p_x(k)*sum(q_row(i)); end PE_theoretical_ML = 1-PC toc Results in the command window: >> HW_DMC_Channel_Estimation_2 S_X = S_Y = p_x_sim = Q_sim = p_y_sim = p_y_sim2 = PE_sim_Naive = PE_theoretical_Naive = Decoder_Table_MAP = PE_sim_MAP = PE_theoretical_MAP =

18 PE_theoretical_MAP = Decoder_Table_ML = PE_sim_ML = PE_theoretical_ML = Elapsed time is seconds. ECS452_HW_2015_2_Sol Page 4

ECS 452: Digital Communication Systems 2015/2 HW 3 Due: Feb 26 Lecturer: Asst. Prof. Dr. Prapun Suksompong Instructions (a) Solve all non-optional problems.

19 ECS 452: Digital Communication Systems 2015/2 HW 3 Due: Feb 26 Lecturer: Asst. Prof. Dr. Prapun Suksompong Instructions (a) Solve all non-optional problems. (5 pt) (i) Write your first name and the last three digit of your student ID on the upper-right corner of every submitted page. (ii) For each part, write your explanation/derivation and answer in the space provided. (b) ONE part of a question will be graded (5 pt). Of course, you do not know which part will be selected; so you should work on all of them. (c) Late submission will be rejected. (d) Write down all the steps that you have done to obtain your answers. You may not get full credit even when your answer is correct without showing how you get your answer. Problem 1 (HW ). Consider a repetition code with a code rate of 1/5. Assume that the code is used with a BSC with a crossover probability p = 0.4. (a) Find the ML detector and its error probability. 3-1

20 ECS 452 HW 3 Due: Feb /2 (b) Suppose the info-bit S is generated with P [S = 0] = 1 P [S = 1] = 0.4. Find the MAP detector and its error probability. (c) Assume the info-bit S is generated with Suppose the receiver observes P [S = 0] = 1 P [S = 1] = (i) What is the probability that 0 was transmitted? (Do not forget that this is a conditional probability. The answer is not 0.45 because we have some extra information from the observed bits at the receiver.) 3-2

21 ECS 452 HW 3 Due: Feb /2 (ii) What is the probability that 1 was transmitted? (iii) Given the observed 01001, which event is more likely, S = 1 was transmitted or S = 0 was transmitted? Does your answer agree with the majority voting rule for decoding? (d) Assume that the source produces source bit S with P [S = 0] = 1 P [S = 1] = p 0. Suppose the receiver observes (i) What is the probability that 0 was transmitted? (ii) What is the probability that 1 was transmitted? 3-3

ECS 452 HW 3 Due: Feb 26 2015/2 (iii) Given the observed 01001, which event is more

Your answer may depend on the value of p 0.

The four possible codewords are 00000, 01000, 10001, and 11111.

22 ECS 452 HW 3 Due: Feb /2 (iii) Given the observed 01001, which event is more likely, S = 1 was transmitted or S = 0 was transmitted? Your answer may depend on the value of p 0. Does your answer agree with the majority voting rule for decoding? Problem 2 (HW ). A channel encoder map blocks of two bits to five-bit (channel) codewords. The four possible codewords are 00000, 01000, 10001, and A codeword is transmitted over the BSC with crossover probability p = 0.1. (a) What is the minimum (Hamming) distance d min among the codewords? (b) Suppose the codeword x = was transmitted. What is the probability that the receiver observes y = at the output of the BSC. 3-4

ECS 452 HW 3 Due: Feb 26 2015/2 (c) Suppose the receiver observes 01001 at the output of the BSC. (i) Assume that all four codewords are equally likely to be transmitted.

23 ECS 452 HW 3 Due: Feb /2 (c) Suppose the receiver observes at the output of the BSC. (i) Assume that all four codewords are equally likely to be transmitted. Given the observed at the receiver, what is the most likely codeword that was transmitted? (ii) Assume that the four codewords are not equally likely. Suppose is transmitted more frequently with probability 0.7. The other three codewords are transmitted with probability 0.1 each. Given the observed at the receiver, what is the most likely codeword that was transmitted? 3-5

ECS 452 HW 3 Due: Feb 26 2015/2 Problem 3 (HW3-2015-2, Optional).

Huffman code for a random variable X has an expected length strictly less than H(X) + 1.

24 ECS 452 HW 3 Due: Feb /2 Problem 3 (HW , Optional). Optimal code lengths that require one bit above entropy: The source coding theorem says that the Huffman code for a random variable X has an expected length strictly less than H(X) + 1. Give an example of a random variable for which the expected length of the Huffman code (without any source extension) is very close to H(X)

27 ECS 452: Digital Communication Systems 2015/2 HW 4 Due: Not Due Lecturer: Asst. Prof. Dr. Prapun Suksompong Problem 1 (HW , Free). In each row of the table below, compare the entropy H(X) of the random variable X in the first column with the entropy H(Y ) of the random variable Y in the third column by writing >, =, or < in the second column. Watch out for approximation and round-off error. H(X) when p = [0.3, 0.7]. H(Y ) when q = [0.8, 0.2]. H(X) when p = [0.3, 0.3, 0.4]. H(Y ) when q = [0.4, 0.3, 0.3]. 0.3, x {1, 2}, H(X) when p (x) = 0.2, x {3, 4}, H(Y ) when q = [0.4, 0.3, 0.3]. 0, otherwise. Problem 2 (HW , Free). Consider random variables X and Y whose joint pmf is given by { c (x + y), x {1, 3} and y {2, 4}, p X,Y (x, y) = 0, otherwise. Evaluate the following quantities. (a) c (b) H(X, Y ) (c) H(X) (d) H(Y ) (e) H(X Y ) (f) H(Y X) (g) I(X; Y ) 4-1

ECS 452 HW 4 Due: Not Due 2015/2 Problem 3 (HW4-2015-2, Free).

28 ECS 452 HW 4 Due: Not Due 2015/2 Problem 3 (HW , Free). Consider a pair of random variables X and Y whose joint pmf is given by 1/15, x = 3, y = 1, 2/15, x = 4, y = 1, p X,Y (x, y) = 4/15, x = 3, y = 3, β, x = 4, y = 3, 0, otherwise. (a) Find the value of the constant β. (b) Are X and Y independent? (c) Evaluate the following quantities. (i) H(X) (ii) H(Y ) (iii) H(X, Y ) (iv) H(X Y ) (v) H(Y X) (vi) I(X; Y ) 4-2

Digital Communication Systems ECS 452

Digital Communication Systems ECS 452 Asst. Prof. Dr. Prapun Suksompong prapun@siit.tu.ac.th Discrete Channel Office Hours: BKD 360-7 Monday 4:00-6:00 Wednesday 4:40-6:00 Noise & Interference Elements