Sporer, ThomasThomasSporerKim, Sang-WokSang-WokKim2025-06-162025-06-162025-03-20https://publica.fraunhofer.de/handle/publica/48863810.71568/dasdaga2025.041In the acoustic community MPEG is recognized as the place where coding standards like mp3, AAC and USAC have been created. These standards exploit properties of the human auditory system to reduce data rate for transmission and storage. Newer standards, like MPEG-H and MPEG-I, improve the perceived quality to 3D and 6DoF. MPEG is now addressing AI algorithms and, in respect to acoustics, machine listening. A key challenge for all AI algorithm is the quality of training data: Bad or insufficient size of training data can lead unexpected behavior of systems (wrong classification, hallucinations). Data used for training must be described as precise as possible with metadata, including description of the content ( what ), recording environment ( where ) and recording configuration ( how ).MPEG started with collecting use cases and requirements in 2022. In the beginning there was a focus on industrial applications, but with more stakeholders becoming involved applications now include ambient assisted living, medical monitoring, and user generated media classification. The scope was extended from training-only to the use of the format for the storage and/or transport of audio data from sensor (arrays) to the classification. The lecture will present the status of the MPEG work including the call for evidence.enAnalyse IndustriegeräuscheMPEGconference paper