Available Topics for Students

We are always looking for skilled students who would like to do a Bachelor or Master thesis in the context of MADMACS. On this page, we present a selection of topics that are currently available. If you discover a topic you are interested in, feel free to contact us. Nevertheless, the list is not complete, so even if your favorite topic is not on it, you may still want to contact us. Also, if you would like to propose your own idea, we are open to hear your suggestions as well.

Multi-Party Interaction with Dynamic Audio Content

In collaboration with HBK Saarbrücken, we are planning to create an interactive multi-person multimedia installation. Rendering audio contents synchronously to multiple mobile devices (based on existing infrastructure) will be one of the main technical aspects of the topic. The content itself (e.g. a game or show) will be another aspect that is designed jointly with HBK. Applicants should have at least basic knowledge of German. See also this flyer.

Implementing a Worker Support Agent with sensor-based Monitoring (Bachelor Thesis)

This topic is about implementing an application for a workshop environment and consists of 2 parts: First, we want to detect based on sensors if a worker is possibly stuck and and may support. Secondly, we want to enable a virtual character to proactively provide support and help avoid frustration.

Frame-base Dialogue Management

The existing SiAM-dp dialogue platform should be extended with a module for the frame-based management of dialogues. Frame-based dialogue queries dynamically for facts rather than following a predefined flow.

Visual / Auditive Multiscale Attention Guidance for Commissioning Environments

Creating an attention guidance scenario (e.g. to a product in a shelf) in a retail store that utilizes different guidance methods based on a person’s distance to the target.

Speech-based Group Detection / Turn-taking

The goal is to detect the group relationship between speakers (which people are in the same group, whom are they currently addressing, and when are they are handing over the word to another person or the system). The primary feature should be the speech metadata (e.g. timing), content (e.g. keywords), and possibly low-level features (e.g. volume, pitch).