Noise pollution is cited as the primary source of discomfort by populations and is a major health and social issue, contributing in particular to stress, sleep disorders, attention deficits, hypertension and tinnitus. Acoustic rehabilitation of existing rooms (e.g. canteens, nurseries) is an important means of remedying this problem. The diagnostic methods used by acousticians today involve a long and costly expertise work, combining on-site sound and geometric measurement protocols with the use of physical simulators and laboratory data. This project aims to radically reduce the time and improve the accuracy of these diagnostics in order to make them accessible to the greatest number of people. The idea is to develop new methods capable of automatically "hearing" the acoustic defects and qualities of rooms from simple "clap" recordings. Hybrid approaches combining signal processing, physical modeling, and supervised learning over hundreds of thousands of simulated environments will be developed to achieve this.
Inria teams involved
In partnership with
UMRAE Strasbourg (UMR d'Univ. G. Eiffel, CEREMA)