In addition to noise detrimental to hearing, the non-auditory effects on cognitive performance and well-being are becoming more relevant in an increasingly digitized workflow. Not only the work type is currently changing, but also the work environment: Many jobs have been moved to the home or other places outside formal workspaces during the covid pandemic. The noise is changing dynamically at these new work locations and there is a need for a flexible form of risk assessment.

This project is intended to develop the foundation for the assessment of noise regarding its effect on mental processes based on subjective judgments. The aim is to explore how machine learning (ML) techniques can support the assessment of noise exposure at low-intensity sound levels. The Technical Rules according to the German Workplace Ordinance have so far provided for a primarily sound pressure level-oriented evaluation. Such an assessment only partially considers the risks from information-containing noise environments, such as irrelevant speech. Noise exposure at low levels can trigger non-auditory noise effects and should be considered with psychoacoustic measures (e.g., noise annoyance). A more recent approach is the annoyance evaluation model (AEM), which introduces the methodology of artificial intelligence to noise assessments. Here, the perceived annoyance is mapped to the properties of the physical sound.

This research project aims to compose realistic soundscapes that can be analyzed and modified for applications as experimental stimuli. The digital composition of soundscapes allows the variation of acoustic parameters. The annoyance of these scenarios should then be assessed in a hearing task with subjects. Using machine learning, a mathematical model is trained regarding the subjective annoyance in the hearing task. To improve assessments of the acoustic working environment it should be determined whether ML systems can provide additional assessment criteria for the implementation in future regulation.

The complete article is published can be downloaded on the website of the "14th ICBEN Congress on Noise as a Public Health Problem" (2023). (charges may apply)