A robot's journey through coupled rooms: "FAIRest Dataset Award" 2024 for Ilmenau scientist

How can acoustic environments be simulated as plausibly as possible for virtual and augmented reality (VR and AR) applications, for example? Georg Stolz is researching this together with his colleagues in the field of virtual acoustics. "On the one hand, we are interested in how sound propagates in a room," explains Georg Stolz. Many factors, such as the room geometry, the interior design or the positions of the sound transmitter and sound receiver in the room, have an influence on this. "On the other hand, we are investigating the human perception of acoustics," says the scientist. In particular, Stolz and his colleagues are trying to find out which differences and changes in room acoustics are audible to people and which are not. One of the aims of this research is to improve the simulation of room acoustics and prediction models of auditory perception. The positions of the sound transmitter and receiver in the room play a major role in this.

One data set for room plus answers at any number of positions in the room

Georg Stolz

To this end, Georg Stolz and his colleagues have equipped a mobile, movable robot platform with a so-called microphone array consisting of seven microphones in order to be able to carry out automated measurements at as many positions in the room as possible. With the help of loudspeakers distributed throughout the room, measurements can be carried out at any number of positions in a room. So-called room impulse responses are measured. "They ultimately represent the influence of the room acoustics on any given audio signal," explains Stolz.

The data set for room impulse responses, for which Georg Stolz and his colleagues have now received an award, was measured in the Helmholtz Building at TU Ilmenau. "The idea came about together with Georg Götz, one of our former students and now a colleague at Aalto University Finland," explains the acoustics expert. The researchers paid particular attention to the so-called acoustic coupling of several rooms and the influence this has on the room acoustics: "So while the robot made its rounds in one of the upper corridors, the sound sources in this data set were distributed over several floors and rooms."

"Subject-specific, open, well documented"

The scientists initially published the finished data set on the European platform for the exchange of scientific data sets Zenodo, where it can be freely downloaded. In addition to the measurements themselves, the research team also took a lot of time to document the data set. This included explanations as well as photos of the setup and overview plans: "We published the measurements as SOFA files (Spatially Oriented Format for Acoustics) so that others could also make use of them. They contain audio data as well as all kinds of metadata on the respective measurement," says Stolz. The aim was to make the data set as easy as possible for other researchers to use, due to its high natural variance also in other fields such as machine learning (ML).

The fact that the researchers used the SOFA format also impressed the judges of the "FAIRest Dataset Award": "It is subject-specific, open, well documented and always has traceability in mind," said the judges in their statement. The work of Georg Stolz and his team therefore meets the principles that are decisive for the presentation of the "FAIRest Dataset Award": Findable, Accessible, Interoperable and Reusable, meaning findable, accessible, interoperable and reusable. By presenting the award at the Thuringian RDM Days 2024, the Thuringian Competence Network for Research Data Management is also recognizing the effort required to make the data available to all interested parties in accordance with these principles. The prize money of 2,000 euros that Georg Stolz and his colleagues receive is earmarked for data management costs.

The Thuringian FDM Days are talking place online this year from June 18 to 19, 2024 under the motto "Becoming literate in the Data Universe - From data literacy to the application of AI".

The data set of the award-winning scientists was created as part of the projects"ISOPERARE", funded by Reality Labs Research at Meta, and "Multiparties", which is funded by the Federal Ministry of Education and Research (BMBF) as part of "KMU-innovativ: Interactive Technologies for Health and Quality of Life":

Stolz, G., Götz, G., Treybig, L., Werner, S., & Klein, F. (2024). Spatial Room Impulse Response Dataset: A Robot's Journey Through Coupled Rooms of a Reverberant University Building [Data set]. DAGA 2024, Hanover. Zenodo. https://doi.org/10.5281/zenodo.10708306