Lecture Course (2.0 VO)
Course Description: This course builds on the fourth semester course Signal Processing and covers adaptive systems in signal processing and control. LMS algorithm and fundamental concepts in adaptive systems: optimality, convergence, stability, accuracy, robustness, tracking, data dependence, computational complexity, implementation and finite word-length effects. Control applications: system idenfitification, self-tuning control, model-reference adaptive control. Signal Processing applications: channel equalization and adaptive detection, echo and noise cancellation, predictive coding and spectral estimation. Selected special algorithms and systems: RLS, blind adaptation, lattice filters, nonlinear systems. The course notes as recorded during lecture presentation in the past terms are available for download in the Teachcenter (login required; link below).
Grading: For the lecture course, the exam consists of a written AND an oral part. For the written part, you have to solve three analytical problems within three hours. These past written exams as well as the problems given below (class material and homework assignments) should help you to prepare. Once you have passed the written exam, you have to take the oral exam (in Prof. Kubin’s office, usually one or two weeks after the written exam). For the oral exam, you have to answer two questions (whiteboard available). You can choose the first question yourself!
Problem Classes (1.0 UE)
In the problem classes we will investigate adaptive systems by using analytical methods as well as numerical simulations in MATLAB. The problem classes should demonstrate typical problems and applications, the theoretical concepts, and state-of-the-art algorithms.
In order to bring the necessary Matlab programming skills, we recommend you to work through the “Getting Started” tutorial (especially the “Manipulating Matrices” section) of the Matlab-Documentation at mathworks.com. It is not mandatory to use Matlab, you can also use Octave as a free alternative which works just as well. If need be, you can use Matlab at home directly over a terminal server. Instructions can be found here.
To pass the problem class you have to deliver 3 homework assignments. For questions and discussion of course-related topics, you can use the newsgroup tu-graz.lv.adaptive. Our assistants will answer newsgroup postings and be responsible for the correction of your homework.
The problems can be found in the Handout (latest version).
|1||11.10.2019||Intro, Least-Squares Filters, Calculus||1.1, 1.2(i,ii)|
|2||18.10.2019||Gradient Calculus, Autocorrelation, MES Filter||1.2(iii,iv), 1.3, 1.4|
Homework Problem Sets
To pass the problem class you have to prepare 3 homework assignments. For each assignment you should work in groups of 2 students (you may change your partner from assignment to assignment). In total, 100 points can be obtained (30 to 35 per assignment). Solving Bonus Problems brings extra points.
|No.||Q&A Date||Due Date||Additional Material|
Discussion of general ideas and questions concerning the homework assignments among students is strongly encouraged. However, all groups are expected to work on their final solutions and documentation individually. Sharing (in particular one-to-one copying) of solutions among groups or copying from other sources such as the internet (including parts of program code) will lead to significant point penalties.
The results of the homework assignments will be posted into the TeachCenter (you need to sign in using your TUGonline account).
How to deliver your homework?
A delayed submission will cost you a penalty of 10 points per day!
The analytical part of the homework (your calculation sheets) as well as the simulation protocol of the MATLAB part have to be delivered as hardcopy to our mailbox at Inffeldgasse 16c / ground floor. You can find a map pointing your way HERE. Use a printed version of the assignment sheets as the title page and fill in your name(s) and matr. number(s). Please make sure that your approaches, procedures and results ara clearly presented.
Additionally, the Matlab/Octave part of the homework (your Matlab/Octave programs and the simulation protocol) has to be submitted via e-mail to the address hw2 dot spsc at tugraz dot at. Please indicate in the simulation protocol wether you used Matlab or Octave (as well as the version number) to avoid any confusion or problems. The subject of the e-mail should be “AssignmentNo MatrNo1 MatrNo2”. Please leave the body of the e-Mail empty (nobody will read it). A complete work consists of all Matlab/Octave files (*.m) and a simulation protocol in PDF format. You have to zip (or tar) all these files to one single file with the name AssignmentNo_MatrNo1_MatNo2 (e.g., “Assignment3_9630815_9833711.zip”), which has to be attached to the e-mail.
- G. Moschytz and M. Hofbauer: Adaptive Filter, Springer-Verlag, Berlin Heidelberg, 2000. (Good intuitive introduction, but available in German only!).
- Simon Haykin: Adaptive Filter Theory, Fourth Edition, Prentice-Hall, Inc., Upper Saddle River, NJ, 2002. (The definitive book but a huge number of pages)
- B. Widrow and S. D. Stearns: Adaptive Signal Processing, Prentice-Hall, Inc., Upper Saddle River, NJ, 1985. (From two of the originators of adaptive system theory. A lot of useful information, but many topics are more detailed covered in Moschytz and Hofbauer)
- J. R. Treichler, C.R. Johnson, and M.G. Larimore: Theory and Design of Adaptive Filters, Prentice-Hall, Inc., Upper Saddle River, NJ, 2001.
- M. G. Bellanger: Adaptive Digitial Filters, Second Edition, Marcel Dekker, Inc., New York, 2001.
Apart from the specific literature mentioned above, we also encourage you to take a look at the following book on signal processing (available for free online here) for any signal processing related issues:
- M. Vetterli, J. Kovačević, and V. K. Goyal: Foundations of signal processing, Cambridge University Press, 2014.
Contest Winners 2015-2016
Congratulations to Gabriel Hülser and Felix Rothmund as the winning team in adaptive systems contest!