Signal Processing and Speech Communication Laboratory
home › phd theses

PhD Theses

Elmar Messner: A Holistic Approach to Multi-channel Lung Sound Classification

Auscultation is the act of listening to the sounds of internal organs. It is an effective non-invasive clinical tool to monitor the respiratory system. Since more than 200 years the stethoscope is used for auscultation. The inherent inter-listener variability and the recent technical advances yield to an increased interest in computer-aided lung sound research.

Martin Trapp: Advances in Learning Sum-Product Networks

Sum-product networks (SPNs) are a recently proposed tractable probabilistic model allowing exact and efficient inference. This thesis focuses on discussing new learning paradigms for SPNs as well as integrating SPNs with recent research in Bayesian nonparameterics. In particular, I focus on the following research questions:

Wolfgang Roth: Bayesian Methods in Machine Learning

Many machine learning algorithms seek for a mode in a posterior distribution and then make predictions solely based on that mode. These approaches waste lots of information available in the posterior and are prone to overfitting when a bad local mode is found. The full Bayesian approach computes its prediction by an expectation over the posterior. Computing this expectation analytically is in general intractable and subject of current research. The task of the thesis is to develop techniques based on the full Bayesian approach:

Stefan Grebien: Cognitive MIMO Radar for RFID Localization

It is hypothesized that the use of cognitive system concepts as well as multiple input, multiple output techniques can enable accurate and robust indoor localization using a passive radio frequency identification (RFID) system.

Martin Ratajczak: Deep Learning and Structured Prediction

Linear-chain conditional random fields (LC-CRFs) have been successfully applied in many structured prediction tasks. LC-CRFs can be extended by different types of deep models.

Johanna Rock: Deep Learning for Resource-Constrained Radar Systems

Radar systems provide information about object distances, velocities and positions but they can also be used for object recognition, such as pedestrian classification. The constant sensing of the various sensors in a car produces a huge amount of raw data, which requires an intelligent extraction of relevant data as well as distributed processing on the sensing units themselves. In this project we investigate the use of machine learning methods such as Deep Neural Networks to address these issues. Special attention is placed on resource efficiency of the used models in order to deploy them directly to the hardware.

Matthias Zöhrer: General Stochastic Networks

Christian Knoll: Numerical Algebraic Geometry for Understanding Machine Learning

Systems of polynomial equations occur in many engineering problems. Finding the common roots of a system of multivariate polynomials is at the heart of various fields of mathematics.

Thomas Wilding: Robust Positioning in Ultra-Wideband Off-Body Channels

The positioning accuracy that can be achieved using communication over a wireless link between transceivers is envisioned to be improved by the introduction of multiple antenna elements at the RF-devices.

Michael Rath: Signal Processing for Localization and Environment Mapping

In the upcoming years we face the reality of everyday objects being connected to a large network, the ‘Internet of Things’ (IoT). A key aspect of the IoT is dependable communication and localization, where the participants act as ‘Smart Things’, communicating with each other and being aware of their environment.

Jamilla Balint: The dynamic of sound fields in enclosed spaces

A basic assumption in room acoustics is that the sound field above the Schröder frequency in an enclosed surface is highly diffuse. But how diffuse can a sound field get? The international standard ISO 354 suggests a measurement procedure to increase the diffusivity of a sound field inside a laboratory environment by adding panel diffusers. Although the standard does not give an absolute measure for diffusivity, it defines the quality of the sound field by measuring the reverberation time and calculating the absorption coefficient of a sample. This procedure is highly questionable because it is not possible to achieve comparable accuracy between different laboratories and the absorption coefficient reaches values > 1 which is physically not possible. In this work, the hypothesis is set up that panel diffusers create coupled spaced and therefore reduce the effective volume of the chamber.

Finished Theses