Short description

We are involved in two projects that are investigating how psychological challenges during space missions can be detected by automatic analysis of communication behavior.We have recorded large amounts of data of group conversations at the Concordia research station in the Antarctica and the AMADEE-18 analog Mars mission in the desert of Oman. The conversation data should be analyzed by a team of psychologists. To make the job easier for them we want to separate the speakers on the recordings and ideally provide them with an automatic transcription. A first step will be to evaluate state-of-the-art algorithms that perform speaker separation. The final system should do automatic evaluation of the communication behaviour, based on parameters extracted.

Tasks (depending on specific project):

• Review of literature and potential algorithms e.g.:FASST
• Choose the most promising algorithm.
• Run the algorithms with the given data and design evaluation apprach
• Implement the most promising approach (Masters’s thesis in)
• Evaluate the chosen approach(es)
• Documentation

Your Profile/Prerequisites

• Motivation and interest in the topic
• Background in Speech Signal Processing
  (e.g. completed Speech Communication 1)
• Experience with either Matlab, Python or Julia
• Can also be done in a team

Contact:

Martin Hagmüller (hagmueller@tugraz.at or 0316/873 4377)