A machine learning technique to identify noise sources in healthcare facilities

What is it about?

Hospital environmental noise can impact the comfort of both workers and patients, causing sleep disturbances and hindering the recovery process. Environmental noise in hospitals is typically due to the mixture of various sound sources, including staff, outdoor noises, and mechanical equipment. Patients and staff experience different effects from noise contributions: the former’s recovery process is influenced, while the latter’s communication efficiency is compromised. Conventional measurements are based on equivalent continuous sound pressure levels. However, the latter does not provide any kind of details about the hospital’s sound context. Unsupervised acoustic measurements have already been used in many fields, but still in a few healthcare scenarios. For this reason, a clustering technique has been used to deepen the conventional noise analysis. The Gaussian Mixture Model (GMM) is used to identify and quantify the contributions of the main types of sound sources. This analysis has been conducted within two four-bed bays: one in active conditions and the other unoccupied. The GMM analysis involves the noisiest period of the day, the morning round. The results demonstrate the GMM’s ability to overcome conventional practices by providing significant insight into environmental noise, even in active conditions. In the active room, noise sources were tagged and measured. The spectral matching, based on the empty room’s outcomes and standard references, represents the benchmark to assess the reliability of the method.

Featured Image

Photo by Adhy Savala on Unsplash

Photo by Adhy Savala on Unsplash