Extending the partial equivalent sound pressure level model to manage residual sound and exclude unwanted events identified with artificial intelligence

Unwanted events, residual sound and irrelevant sounds affect the quality of measurement results obtained by long-term noise monitoring. Previous studies have shown that irrelevant sounds can be identified with direction-of-arrival techniques, which allow using corrected data in a model called partial equivalent sound pressure level. Under given assumptions, the model can also account for residual sound. Other studies have shown that artificial intelligence can be used to automatically identify unwanted events in noise monitoring measurements. In this paper, we extend the partial equivalent sound pressure level model by correcting for unwanted events, in addition to the previously included corrections for residual sound and irrelevant sounds. The proposed extended model is applied to a case study, which is presented together with a discussion of the assumptions made and the limitations of the method. We compare the results with those achieved by applying the methods given in ISO 1996-2. Our primary hypothesis is that better estimates of equivalent sound pressure levels can be achieved, but other factors such as the efficiency and reliability of the extended model are also analyzed. Finally, further work towards a generalization of the method is mentioned.