Lectures

Winter Term 2024/25

Lectures

Signals und Systems I

  • Content: This lecture covers an introduction to the theory of continuous signals and continuous linear time-invariant (LTI) systems. At the beginning, elementary signals, the delta impulse, the convolution, and the correlation of signals are discussed. Subsequent to this, the frequency domain representation of signals by means of Fourier and Laplace transform is introduced, including the theorems and correspondences of these transforms. Next, the time-domain description of LTI systems by impulse response and convolution, differential equations, and state-space representation is presented. In addition to this, the description of systems in the frequency domain using eigen functions, the transfer function, the system function, and state-space representation is given. Furthermore, LTI systems with initial conditions are considered. Following, after linear phase, minimum phase, idealized, and all-pass system have been introduced, causality, the Hilbert transform, stability and feedback systems are discussed. The lecture closes with an overview of sampling systems and the sampling theorem for lowpass and bandpass signals.
  • Lecturer: André Kaup
  • Exercise: Frank Sippel
  • Tutorium: Felix Deichsel
  • ECTS: 5.0
  • SWS: 3.0
  • Language: German
  • More information can be found on StudOn and Campo.

Signal Theory

  • Content: This lecture covers an introduction to the theory of continuous signals and continuous linear time-invariant (LTI) systems. At the beginning, elementary signals, the delta impulse, the convolution, and the correlation of signals are discussed. Subsequent to this, the frequency domain representation of signals by means of Fourier and Laplace transform is introduced, including the theorems and correspondences of these transforms. Next, the time-domain description of LTI systems by impulse response and convolution, differential equations, and state-space representation is presented. In addition to this, the description of systems in the frequency domain using eigen functions, the transfer function, the system function, and state-space representation is given. Furthermore, LTI systems with initial conditions are considered. Following, after linear phase, minimum phase, idealized, and all-pass system have been introduced, causality, the Hilbert transform, stability and feedback systems are discussed. The lecture closes with an overview of sampling systems and the sampling theorem for lowpass and bandpass signals.
  • Lecturer: Jürgen Seiler
  • Exercise: Teresa Stürzenhfäcker
  • Tutorium: Felix Deichsel
  • ECTS: 5.0
  • SWS: 3.0
  • Language: English
  • More information can be found on StudOn and Campo.

Image, Video, and Multidimensional Signal Processing

  • Content: The lecture gives an introduction to the basics of image and video signal processing. First, point operations, morphological filters, and color spaces including tri-chromaticity are explained. Subsequently, the theory of multidimensional signals and systems is introduced and Wiener filtering for image signals is derived. Based on this, interpolation methods for images such as bicubic and spline interpolation are explained. This is followed by methods for feature detection in images using Hough transforms and edge detection, and the principle of scale-invariant features is explained. For video signals, motion estimation methods such as optical flow and image matching algorithms using SIFT and SURF are explained. Finally, the theory of image and Video segmentation using statistical methods is introduced, and transform-based methods for image processing are presented.
  • Lecturer: André Kaup
  • Exercise: Andy Regensky
  • ECTS: 5.0
  • SWS: 3.0
  • Language: German
  • More information can be found on StudOn and Campo.

Communication Networks

  • Content: The lecture gives an introduction to fundamental concepts and mechanisms of digital communication networks. After explanation of some basic terms we first introduce hierarchical structuring of networks leading to the OSI reference model. Subsequent to discussion of data transmission from point to point, data link protocols for errorfree transmission are discussed, especially forward error control and ARQ protocols. Protocols for multiple access control follow, among them ALOHA protocols, strategies to detect and resolve collisions, carrier sensing, as well as token passing principles. Subsequently, fundamental routing algorithms are explained. After an introduction into queuing theory, the lecture gives an overview of the internet protocol suite TCP/IP as important system example and closes with an analysis of multimedia networks.
  • Lecturer: André Kaup
  • Exercise: Matthias Kränzler
  • ECTS: 5.0
  • SWS: 3.0
  • Language: English
  • More information can be found on StudOn and Campo.

Digital Signal Processing

  • Content: The course assumes familiarity with basic theory of discrete-time deterministic signals and linear systems and extends this by a discussion of the properties of idealized and causal, realizable systems (e.g., lowpass, Hilbert transformer) and corresponding representations in the time domain, frequency domain, and z-domain. Thereupon, design methods for recursive and nonrecursive digital filters are discussed. Recursive systems with prescribed frequency-domain properties are obtained by using design methods for Butterworth filters, Chebyshev filters, and elliptic filters borrowed from analog filter design. Impulse-invariant transform and the Prony-method are representatives of the considered designs with prescribed time-domain behaviour.For nonrecursive systems, we consider the Fourier approximation in its original and its modified form introducing a broad selection of windowing functions. Moreover, the equiripple approximation is introduced based on the Remez-exchange algorithm. Another section is dedicated to the Discrete Fourier Transform (DFT) and the algorithms for its fast realizations (‘Fast Fourier Transform’). As related transforms we introduce cosine and sine transforms. This is followed by a section on nonparametric spectrum estimation. Multirate systems and their efficient realization as polyphase structures form the basis for describing analysis/synthesis filter banks and discussing their applications. The last section is dedicated to investigating effects of finite wordlength as they are unavoidable in any realization of digital signal processing systems. A corresponding lab course on DSP will be offered in the winter term.
  • Lecturer: Heinrich Löllmann
  • Exercise: Heinrich Löllmann
  • ECTS: 5.0
  • SWS: 3.0
  • Language: English
  • More information can be found on StudOn and Campo.

Statistical Signal Processing

  • Content: 
    • Discrete-Time random processes in time domain and frequency domain
    • Random variables (RVs): probability distributions and densities, expectations; transformation of RVs; vectors of normally distributed RVs; discrete-time random processes (RPs): probability distributions and densities, expectations; stationarity, cyclostationarity, ergodicity, correlation functions and matrices, frequency-domain representations; principal component analysis, Karhunen-Loeve transform;
    • Estimation theory
    • Estimation criteria; prediction; classical and Bayes estimation (incl. MMSE, Maximum Likelihood, Maximum A Posteriori); Cramer-Rao bound
    • Linear signal models
    • Nonparametric models (cepstral decomposition, Paley-Wiener theorem, spectral flatness); Parametric models: allpole /allzero/pole-zero-(AR/MA/ARMA) models; lattice structures, Yule-Walker equations, PARCOR coefficients, cepstral representation;
    • Signal estimation
    • Supervised signal estimation, problem classfication; orthogonality principle, MMSE estimation, linear MMSE estimation for normal processes; optimum FIR filters; optimum lineare filters for stationary RPs; prediction, filtering, smoothing; Kalman filter; optimum multichannel filters (Wiener filter, LCMV, MVDR, GSC);
    • Adaptive Filtering
    • Gradient descent; LMS, NLMS , APA and RLS algorithm and its convergence
  • Lecturer: Sebastian Schlecht
  • Exercise: Jeremy Baoqi Bai
  • ECTS: 5.0
  • SWS: 3.0
  • Language: English
  •  More information can be found on StudOn and Campo.

Machine Learning in Signal Processing

  • Content: The students will learn the basics of machine learning, including theory and applications. The students will acquire knowledge on processing different types of data and signals with machine learning algorithms. In the exercises, the students will work on regression and classification assignments as well as different machine learning approaches.

    The lecture topics include:

    • Machine learning basics
    • Linear regression
    • Linear classification
    • Performance metrics
    • Decision trees
    • Random forests
    • Support vector machines
    • Neural networks
    • Clustering
  • Lecturer: Vasileios Belagiannis
  • Exercise: Amir El-Ghoussani
  • ECTS: 5.0
  • SWS: 3.0
  • Language: English
  • More information can be found on StudOn and Campo.

Introduction in Deep Learning

  • Content: The students will learn the basics in deep learning, including classical neural network models and recent architectures. The students will acquire knowledge on processing different types of data with deep neural networks. In the exercises, the students will implement some of the standard models for classification or regression tasks and acquire knowledge on machine learning applications.

    The lecture topics include:

    • Learning from data, machine learning and deep learning
    • Machine learning principles
    • Artificial neural networks
    • Convolutional neural networks
    • Back-propagation
    • Network optimization
    • Initialisation, regularisation
    • Deep network architectures
    • Generative models
    • Auto-encoders
    • Sequential models
    • Deep learning applications
  • Lecturer: Vasileios Belagiannis
  • Exercise: Marc Hölle
  • ECTS: 5
  • SWS: 4
  • Language: English
  • More information can be found on StudOn and Campo.

Music Processing - Synthesis

  • Content: The lecture covers digital processing of audio signals with parametric filters and effects as well as sound synthesis for musical applications. Various sound examples demonstrate the contents of the lecture. *Filters and Effects* – parametric filters: structure and design – digital sound effects *Digital Sound Synthesis* – a short history of computer music – wavetable synthesis – spectral synthesis – physical modelling *Systems for Sound Production and Reproduction* – sound effects – synthesizer and vocoder – artificial reverberation
  • Lecturer: Maximilian Schäfer
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • More information can be found on StudOn and Campo.

Signal Analysis

  • Content: In this course, different approaches for the analysis of digital signals and their applications are treated, which comprises the following topics:
    • Fourier analysis of signals
    • Signal analysis by means of time-frequency transformations
    • Parametric and non-parametric signal analysis
    • Frequency estimation
    • Spatial signal analysis
    • Filter-banks and wavelets.
  • Lecturer: Heinrich Löllmann
  • ECTS: 5.0
  • SWS: 2.0
  • Language: English
  • More information can be found on StudOn and Campo.

Seminars

Selected Topics in Multimedia Communications and Signal Processing

  • Content: This seminar is designed for Bachelor and Master programs in Electrical Engineering, Electronics and Information Technology (EEI), Information and Communication Technology (IuK), Communications and Multimedia Engineering (CME), Advanced Signal Processing and Communications Engineering (ASC) as well as related study programs.
    The students will study, understand, and present scientific publications from the recent literature on applied artificial intelligence, machine learning and deep learning. After finishing the seminar, the student will be able to summarise and present a publication.
  • Organizer: Vasileios Belagiannis
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • More information can be found on StudOn and Campo.

Lab courses

Image and Video Signal Processing on Embedded Systems

  • Content: Today, many image and video signal processing applications are running on embedded systems. However, the computational power and the energy storage is a limiting demand for embedded systems. Nevertheless, daily mobile devices like smartphone and tablet are able to perform signal processing tasks for image and video signals, for example coding of images and videos, the creation of a panorama or the calculation of images with high dynamic range. The image and video signal processing on embedded systems lab course should show the challenges that occur while handling with such mobile devices and the implementation of such algorithm on an embedded system. Therefore, Raspberry Pis as embedded systems and Python as coding language is used in the laboratory. The experiments include the setup of the Raspberry Pi, an introduction to Python and an introduction to image and video signal processing. In addition, a camera will be connected, signal processing will be done with the camera and digital filters are implemented. Moreover, the laboratory includes different computer vision applications like the creation of a panorama.
  • Organizers: Jürgen Seiler, Thanh Dat Nguyen
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • More information can be found on StudOn and Campo.

Digital Signal Processing

  • Content: In this laboratory course the theory from the lecture Digital Signal Processing is applied in practice, using the programming environment MATLAB. The topics include quantization, spectral analysis, FIR and IIR filter design, filter banks and adaptive filters. The course consists of 5 guided experiments in which students work on programming problems in groups of two, and a group project in which each group works on an individual project from the field of digital signal processing. The preparation, as well as the results of the past experiment will be examined by a short test at the beginning of each experiment. For passing the lab course, a minimum number of points from the tests and the project is required. The course requires previous experience in MATLAB programming. It is possible take the course in parallel to the DSP lecture, however, revision of the relevant lecture contents before each lab lesson, and participation in the DSP exercises and tutorials is required.
  • Organizers: Matthias Kreuzer, Heinrich Löllmann
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • More information can be found on StudOn and Campo.

Lab Course Maschine Learning and Systems

  • Content: The students will learn to develop machine learning algorithms for systems. Lab projects will focus on efficient model training and inference, hardware-aware algorithms, interpretability, and robustness of machine learning systems. Deep neural networks will be the main approach for development. The assignments will include tasks such as:
    • Neural network compression.
    • Machine learning algorithms on embedded devices.
    • Automated driving applications.
    • Generative models.
    • Model interpretability.
    • Benchmarking.e
  • Organizers: Michele De Vita, Vasileios Belagiannis
  • ECTS: 2,5
  • SWS: 2.0
  • Language: English
  •  More information can be found on StudOn and Campo.

Simulation-Tools

  • Content: This lab course provides an introduction to the programming language MATLAB and is part of the teaching event “Arbeitstechnik”. The course is conducted online and consists of 4 experiments:
    • Introduction to MATLAB
    • Graphics in MATLAB
    • Matrices and linear equations
    • Complex Numbers.
  • Organizer: Heinrich Löllmann
  • SWS: 2.0
  • Language: German
  •  More information can be found on StudOn and Campo.

Summer Term 2024

Lectures

Signals and Systems II

  • Content: The lecture covers the following topics:
    • Discrete signals
    • Discrete-time Fourier transform (DTFT)
    • Discrete Fourier transform (DFT)
    • z-transform
    • Discrete LTI systems in the time domain
    • Discrete LTI systems in the frequency domain
    • Discrete LTI systems with special transfer functions
    • Causal discrete LTI systems and Hilbert transform
    • Stability of discrete LTI systems
    • Description of stochastic signals
    • Stochastic signals and LTI systems
  • Lecturer: André Kaup
  • Exercise: Simon Deniffel
  • Tutorium: Hannah Och
  • ECTS: 5.0
  • SWS: 3.0
  • Language: German
  • Further information can be found on StudOn and Campo.

Image and Video Compression

  • Content: The lecture gives an introduction to basic concepts and algorithms for coding and transmission of image and video signals. We start with sampling and digital representation of images and video signals and learn about fundamental features of human visual perception. Principles of data compression using redundancy and irrelevancy reduction are discussed in detail together with typical algorithms used for coding of image and video signals. Among others this covers the design of quantizers with the help of the Lloyd-Max algorithm, entropy coding using Huffman and arithmetic coding, as well as run length coding. Moreover, fundamentals of vector quantization and predictive coding are addressed. Methods for frequency analysis will be discussed taking transform coding, subband decomposition and wavelet analysis as examples. The principle of motion compensation and hybrid video coding is introduced. Finally, current MPEG and ITU standards for coding of still images and video signals are explained.
  • Lecturer: André Kaup
  • Exercise: Anna Meyer
  • ECTS: 5.0
  • SWS: 3.0
  • Language: English
  • Further information can be found on StudOn and Campo.

Stochastische Prozesse

  • Content:
    • Calculus of probabilities and random variables
      probability, random variables, uni- and multivariate probability density functions and cumulative distribution functions; functions of random variables and their distributions and densities; expected values; special distributions (discrete and continuous); limit theorems
    • Stochastic processes
      distributions, densities and expected values of one-dimensional stochastic processes; stationarity, cyclostationarity, ergodicity; weakly stationary, continuous-time and discrete-time processes in the time domain and in the frequency domain; linear time-invariant (LTI) systems and weakly stationary processes

    • Estimation theory
      point estimation and interval estimation; estimation criteria; prediction; classical and Bayesian parameter estimation (incl. MMSE, Maximum Likelihood, Maximum A Posteriori); Cramer-Rao bound; statistical hypothesis tests and decision processes (binary decisions, test statistics, Chi-Squared test); binary decisions, Neyman-Pearson criterion

    • Linear optimal filtering
      principle of othogonality; continuous-time and discrete-time Wiener filtering; adaptive filters (LMS, NLMS); continuous-time and discrete-time matched filter

  • Lecturer: Sebastian Schlecht
  • Exercise: Sebastian Schlecht
  • ECTS: 5.0
  • SWS: 3.0
  • Language: German
  • Further information can be found on StudOn and Campo.

Transformations in Signal Processing

  • Content: The lecture “Transformations in Signal Processing” covers several different transforms which are used in the field of signal processing. For this, first the basic concepts of transforms are discussed and the advantages which are offered by the different transforms are presented. Subsequent to this, fundamental properties of integral transforms are considered and the Laplace- and the Fourier-Transform are examined in detail. To be able to transform time-varying signals, the Short-Time Fourier-Transform and the Gabor-Transform are introduced, afterwards. Subsequent to this, the impact of sampling on transformed signals is analyzed before the z-Transform as a transform for discrete signals is covered. Finally, further transforms for discrete signals like the Discrete Fourier-Transform or Linear-Block Transforms are discussed.
  • Lecturer: Jürgen Seiler
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • Further information can be found on StudOn and Campo.

Introduction in Deep Learning

  • Content: The students will learn the basics in deep learning, including classical neural network models and recent architectures. The students will acquire knowledge on processing different types of data with deep neural networks. In the exercises, the students will implement some of the standard models for classification or regression tasks and acquire knowledge on machine learning applications.

    The lecture topics include:

    • Learning from data, machine learning and deep learning
    • Machine learning principles
    • Artificial neural networks
    • Convolutional neural networks
    • Back-propagation
    • Network optimization
    • Initialisation, regularisation
    • Deep network architectures
    • Generative models
    • Auto-encoders
    • Sequential models
    • Deep learning applications
  • Lecturer: Vasileios Belagiannis
  • ECTS: 5
  • SWS: 4
  • Language: English
  • More information can be found on StudOn and Campo.

Advanced Topics in Deep Learning

  • Content: The students will learn advanced deep learning topics, including recent network architectures, generative models, self-supervision, interpretability and explainability. In the exercises, the students will implement advanced models and techniques for classification or regression tasks.The lecture topics include:
    • Geometric deep learning
    • Attention and transformers
    • Unsupervised and self-supervised learning
    • Generative models
    • Interpretability
    • Explainability
    • Efficient Inference
    • Uncertainty estimation
    • Transfer learning and domain adaptation
    • Few-shot learning
  • Lecturer: Vasileios Belagiannis
  • ECTS: 5
  • SWS: 4
  • Language: English
  • More information can be found on StudOn and Campo.

Perception in Robotics

  • Content: The students will learn robotic perception topics, including camera models, filtering, transformations, low-level features, point-cloud processing, recognition, pose estimation, localization, mapping, depth, and motion estimation. In the exercises, the students will implement techniques for different perception modules.The lecture topics include:
    • Sensor models
    • Camera calibration
    • Feature detection and matching
    • Edges, lines, circles
    • Transformations
    • Multiple views
    • Recognition
    • Pose estimation
    • Localization and mapping
    • Depth estimation
    • Point-cloud processing.
  • Lecturer: Vasileios Belagiannis
  • Exercise: Michele De Vita
  • ECTS: 5
  • SWS: 4
  • Language: English
  • More information can be found on StudOn and Campo.

Seminars

Selected Topics of Multimedia Communications and Signal Processing

  • Content: The seminar deals with current research topics in the area of multimedia communications and signal processing. In an introductory meeting, the course of the seminar is outlined and each participant selects one of the offered topics. The participant should become familiar with the assigned research topic and present it by a report and a talk at the end of the seminar with the support of a supervisor.
  • Organizer: Vasileios Belagiannis
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • Further information can be found on StudOn and Campo.

Selected Topics in Machine Learning

  • Content: The seminar deals with current research topics in the area of Machine Learning and Deep Learning. In an introductory meeting, the course of the seminar is outlined and each participant selects one of the offered topics. The participant should become familiar with the assigned research topic and present it by a report and a talk at the end of the seminar with the support of a supervisor.
  • Organizer: Vasileios Belagiannis
  • ECTS: 2.5
  • SWS: 2.0
  • Language: English
  • Further information can be found on StudOn and Campo.

Deep Learning in Image and Video Processing

Lab courses

Image and Video Compression

  • Content: The lab course focuses on video coding methods. During the lab course, each group will implement its own video codec and evaluate its properties. Knowledge in MATLAB is helpful but not required.

    Structure
    The lab course consists of four main parts:
    Part 1: Introduction to MATLAB
    Part 2: Processing blocks in video codecs (quantization, entropy coding, transformation, motion estimation)
    Part 3: Integration of the blocks into a video codec pipeline and implementation of optional methods
    Part 4: Subjective evaluation and analysis

    Additional information
    The content of this lab course is closely connected to the lecture Image and Video Compression (IVC).

  • Organizer: Geetha Ramasubbu
  • ECTS: 2.5
  • SWS: 3.0
  • Language: English
  • Further information can be found on StudOn and Campo.

Machine Learning in Signal Processing