Cognitive Neuroinformatics

Kameras haben meist Millionen von Pixeln, von denen ein paar defekt sein können und falsche Helligkeitswerte liefern. Bereits ein winziger Bruchteil an defekten Pixeln im Bild fällt einem Betrachter sehr unangenehm auf und muss korrigiert werden. Die bekannten Verfahren gehen dabei oft viel zu aggressiv vor. Sie korrigieren auch tausende guter Pixel und radieren dabei interessante Detailinformationen aus dem Bild heraus.

In this thesis, you will focus on the application and implementation of data mining methods to discover dependencies in medical data. In a series of experiments, we will non-invasively gather physiological data of people who are diagnosed with apallic syndrome (who are in a persistent vegetative state). Though the application of suitable data mining strategies we want to become able to make reliable statements on the cognitive abilities of the patients, in order to support medical staff and caregivers in their decision-making process.

If you are interested and want to talk about further details, please contact Carolin Zschippig.

Diese Abschlussarbeit fällt in das Gebiet der Steuerung dynamischer Systeme. Es soll eine Programmierschnittstelle für einen Roboterarm geschrieben und ein Modell des Arms entwickelt werden, so dass eine optimale Steuerung des Greifers entlang eines 3D-Pfads (berücksichtigt Hindernisse) möglich ist.

The recognition of a place from visual input is a challenging problem. In the last decade there has been significant progress in this field.

This project focuses on building a bio-inspired scene analysis system which is able to infer a place from one or few snapshots. It will be implemented on a "robot head", which enables the system to change his viewing direction by movements of the head and the cameras. Different machine vision/patter recognition techniques which relate to properties of the human visual system will be investigated and combined:

• foveal vision: extraction of visual features from salient/informative regions
• peripheral vision: e.g. extracting the distribution of low selective features in the whole visual field
  
• active vision: top-down selection of regions to be analyzed

The diploma or master thesis will focus on selected aspects and can build upon previous work from our group. Some previous knowledge in machine vision and digital signal processing is helpful but not required.