EE640 SPRING 2005
STOCHASTIC SYSTEMS
INFORMATION
Updated 4-1-05
VISUALIZAIONS (not updated from 2004 yet)
Example of Stationary Colored Noise versus a non-stationary image, both having same PSD
Data Whitening and the Covariance Matrix
PROJECTS
PROJECT 1a:
PART A: SYNTHESIS (Prjct1a2004.doc)
PROJECT 1b:
PART B: ANALYSIS (Prjct1b2005.doc)
PROJECT 1c:
PART C: DETECTION AND DISCRIMINATION (Prjct1c2005.doc)
Journal References for project 1
IF STUDENTS ELECT TO HAVE A PROJECT INSTEAD OF A FINAL EXAM, THE PROJECT WILL BE THE FOLLOWING
PROJECT 2a:
ENCODING INFORMATION INTO THE COLOR NOISE OF A RGB IMAGE (ee640project2a03.doc)
DIGITAL COMMUNICATIONS BELOW THE NOISE FLOOR (see below for instructions)
PROJECT 2b :
PROTOCOL
for PROJECT 2B: updated 4-28-03
Student sends the bit matrix size, modulator and
demodulator m files to instructor. All the files sent to the instructor have the
"groupname" as a prefix so the instructor can keep the track of the
individual group m files and data.
1. BIT MATRIX SIZE (student sends this to instructor): The
student is ranked by the total number of bits that can be transmitted through
the channel. The bit matrix is 2 dimensional. It has a length Nbit (column
dimension) and a width of Nseq (row dimension). These values, named
"Nbit" and "Nseq", along with a character string containing
"groupname", they are stored in a file called
"groupname_Bsize.mat." An example m file is "groupname_createBsize.m".
Group sends instructor this m file to initiate test.
2. BIT MATRIX (instructor generates this based on
groupnameBsize values): The bit matrix is generated and stored in a file called
groupname_B.mat and the matrix is called B. A sample code that will generate a
Nseq x Nbit bit matrix B is "Bgen.m."
I will use Bgen to generate a random sequence of bits of the size specified by
the student in "groupnameBsize.mat."
3. MODULATOR (student sends the modulator m file to the
instructor): A modulator m file by the name "groupname_modulator.m"
will be sent to the instructor. Its input is the file named
"groupname_B.mat." The program will create a 1 x N real vector
"s" and a Nseq x N, bit check matrix, called "Bcheck." The
signal vector will be stored in "groupname_signal.mat" and the bit
check matrix is stored in "groupname_Bcheck.mat." The length is
N=524288=65536*8. The Bcheck matrix (Nseq x N) has 3 element values +1 for a
bit value of "1" to be present, -1 for a bit value of "0"
to be present and 0 for "don't care."
4. CHANNEL (instructor will run this program, channel.m, on
vector s). The channel will do two things, lowpass filter and then add noise.
The signal vector s0 is convolved with the Butterworth low pass filter of order
8 and fc=N/8, yielding a bandlimited signal vector s. The noise is based on the
value sigma=2*(max(s)-min(s)) and is generated by w=sigma*randn(1,N). The noisy
vector is sn=s+w. The output of the channel will be a real one dimensional
vector, "r", of size 1xN. This r vector will be stored in
groupname_r.mat.
5. DEMODULATOR/BINARIZER (student sends the demodulator.m
file to the instructor): A demodulator file by the name "groupname_demodulator.m"
will be sent to the instructor. Its input file is groupname_r.mat. Its
output will be a Nseq x N real matrix. Each row of the matrix will represent
the demodulated and binarized bit stream defined in B. This output will be
stored in "Bs" and saved to the file groupname_Bs.mat. NOTE: The
demodulator should also binarize the signals in Bs to have values of either 1
or 0.
6. BIT CHECK (instructor will run bitcheck, bitcheck.m, to
test the students data for errors): The instructor will run a program that will
input the groupname_B.mat file, groupname_Bcheck.mat file and the
groupname_Bs.mat file. The program will go to each value of 1 or -1 in the
Bcheck matrix and see if the associated element in the Bs matrix is (1) if
Bcheck is 1, then Bs value must be 1 (above 0.5), (2) if Bcheck is -1, then Bs
must be 0 (below 0.5). For each element of Bcheck that is 0 (between -0.5 and
+0.5) the associated value in Bs is ignored. The resulting values will be
verified with the B matrix. To be acceptable, there must not be any errors in
either the number of ones and zeros or the specific bit values when compared to
B. The results will be posted on the web.
Additional m files include:
