Information Processing

A tantárgy neve magyarul / Name of the subject in Hungarian: Információfeldolgozás

Last updated: 2012. november 25.

Budapest University of Technology and Economics
Faculty of Electrical Engineering and Informatics
Electrical Engineering, MSc course
Embedded Systems branch
Course ID Semester Assessment Credit Tantárgyfélév
VIMIM237 2 2/1/0/v 4  
3. Course coordinator and department Dr. Kollár István,
Web page of the course http://www.mit.bme.hu/oktatas/targyak/vimmm237/
4. Instructors

Prof. István Kollár

6. Pre-requisites
Kötelező:
NEM ( TárgyEredmény( "BMEVIMIMA10" , "jegy" , _ ) >= 2
VAGY
TárgyEredmény("BMEVIMIMA10", "FELVETEL", AktualisFelev()) > 0)

A fenti forma a Neptun sajátja, ezen technikai okokból nem változtattunk.

A kötelező előtanulmányi rendek grafikus formában itt láthatók.

7. Objectives, learning outcomes and obtained knowledge

This subject deals with characterization, extraction and complex processsing of information (measured signals, measured quantities, etc.), collected about the surrounding world. Physical quantities are related to the quantities stored in the computer, possibilities of information extraction are discussed. In relation to embedded systems, fast methods of partial information extraction are also treated. These methods are sometimes autonomous, sometimes human controlled by humans.

Students accomplished this subject should be

  • able to evaluate the information included and extractable from the measured signals,
  • aware of basic engineering descriptions of signals and systems, methods of modelling,
  • capable to use basic computer-based methods of information extraction,
  • able to analyse existing systems, by examining modelling and representation errors, efficiency of information extraction, run time, etc.,
  • capable to design such systems
  • able to understand, handle and use information from heterogenious sensor systems.
8. Synopsis

I Fundamentals of information extraction and system modeling (6 weeks)

Model fitting, relation of computer model and reality. Model types (disckrete-time, continuous-time, deterministic and stochastic, etc.). Analogy  of differennt physical phenomena (same difference equation).

Stochastic processes. Stationarity and ergodicity. Examples. Disrete Fourier Transform: properties of the DFT of randomly-timed periodic signals. Processes with continuous power spectral density.

Sampling, quantization, roudoff, dither. Theorems and practical applicability of these. Systems containing ADC's. Matching differently samples sequences, system design. Examples and counterexamples.

Averaging. Relation of discrete and continuous data processing. Averaging and lowpass filtering. Reconstruction of the continuous-time signal from samples. Explicite and recursive averaging, stability. Signal enhancement and moving average.

Basic quantities used in signal processing. Compression. Correlation function. Power spectral density. Periodogram and circular correlation. Spectral analysis with bandpass filters. Filter banks. Real and complex modulation, zoom, On-line signal processing.

Embedded systems, modeling and system identification. Network analysis. Measurement and experiment design. Excitation signals: multisine and sweep sine, noise, pseudorandom noise and impulse/step response.

II. Qualitative and knowledge-intensive methods of information processing

Machine learning (2 weeks)

      Learning and adaptivity. Model of the learning problem.

      Learning by examples. Learning theory.

     Supervised logics learning. Decision trees.

     Artificial neural networks. Backproparagion learning.

     Examples and demonstrations.

     Learning based on validation or decision value.

Probability nets (1 week)

     Network representation of probabilistic information processing.

     Data processing. Logical sampling.

     Learning of probabilities by examples. Diagnosis with nets.

Rule-based systems (1 week)

     Illustration and manipulation of information.

     Rule based systems. Forward and backward reasoning.

     speedup of rule-based systems. Implementations.

Fuzzy logic methods (1 week)

     Basics of fuzzy logic. Fuzzy sets and membership functions.

     Fuzzy inference.

     Typical fuzzy information processing schemes. Fuzzy signal processing. Fuzzy coontrollers, fuzzy radial functions.

III. Sensor fusion (1 week)

Levels of sensor fusion. Typical problems.

Fusion at signal processing level.

Consensus-based filters.

Fusion with neural and probabilistic nets.

Dempster-Shaffer theory, fusion with fuzzy logic.

Complex and hybrid examples.

Material for self-study:

Several details of sampling

Signal processing in the time domain, ad hoc methods, partial information extraction with fast algorithms. Measurement of rise time, delay, peak value.

Design of digital filters.

9. Method of instruction Lectures
14. Required learning hours and assignment
Lessons
 42
Preparation for the lectures
 10
Preparation for the test
 15
Homework 8
Independent studying
 15
Preparation for the examination
 30
Altogether 120