Analyzing Fluctuating Soundscapes Using Machine Learning to Better Understand Occupant Experience

Multisource, time varying acoustic environments, such as hospitals and early childcare centers, require in-depth analyses to understand the impacts sound has on its occupants. The hospital soundscape can have a profound impact on its occupants, such as reduced sleep and extended stays for patients and reduced performance and job satisfaction for staff. While less is known about early childcare soundscapes, out-of-home childcare is a normative experience during a critical period of development and learning. Both these environments have the potential to benefit from insights into occupant soundscape perception that can be provided through advanced statistical methods such as machine learning. In the hospital phase of this research, 9 hospital patient rooms were acoustically measured for 24-hours, a staff soundscape perception survey was administered, and patient satisfaction was measured through a national survey. In the early childcare phase of this research, 15 rooms in three early childcare centers were acoustically measured for two weekdays, a staff follow-up survey was administered, and child language data was collected. In both phases, unsupervised clustering machine learning was utilized to extract patterns of activity in background noise and calculate novel activity metrics which were correlated to occupant outcomes of hospital patient satisfaction and child language data. Results indicated acoustic activity conditions calculated through clustering were significantly related to occupant outcomes. Overall, the benefit and need of a thorough, well-rounded approach to analyzing these complex acoustic environments was realized