Course Description

EN2500/FEN3500

Period 3, 7.5/12 ECTS credits

The course treats the principles and methods used to reduce the bit rate required to describe signals such as video, audio, speech, and images. The emphasis is on the underlying theoretical principles. The course covers the information theory of discrete and continuous variables, the estimation of probability distributions from data, lossless coding, rate-distortion theory, high-rate quantization theory, the generalized Lloyd algorithm for quantizer design, and transforms and signal-modeling techniques in the context of coding.

The course is given in English.

The course is an Accelerated Program Course (difficult but counts towards PhD at the KTH School of Electrical Engineering).

Graduate students: you formally take this course with another code, FEN3500, but different is only that there is an optional project part (additional 4.5 ECTS credits). If you only take the lecture course then FEN3500 and this course are identical in all aspects. Some suggested topics for optional project include but not limited to: Other bounds on the rate-distortion function; Multiple-description coding; Aspects of lattice quantization (faster and better search procedures or the usage of specific lattices); Alternative training procedures for vector quantization; Low-rate coding using significance map; Matching pursuit techniques; Tree structured coding; Embedded coding and successive refinement.

Prerequisites

EQ1220 Signal Theory or equivalent.

Course literature

T.M. Cover and J.A. Thomas, “Elements of Information Theory,” John Wiley & Sons, Inc., New York.

W.B. Kleijn, A basis for source coding.