Development of a sound quality model for noise impact prediction in building acoustics

Abstract

Building acoustics is the science of controlling and predicting sounds in buildings, as well as studying the effects of environmental sound quality and noise impacts on building occupants. However, developing noise impact prediction remains a research challenge due to a huge variety in multidimensionally modeling of environmental sound quality. Researchers have recently revealed the three fundamental human perceptual dimensions of sound, called Evaluation, Potency, and Activity (EPA). This is the first study aimed at developing prediction models for the negative noise impacts (O1: Discomfortable, O2: Annoying, O3: Stressful, O4: Unacceptable, O5: Averse, O6: Unsatisfying) using the sound quality model (EPA model) in three approaches (holistic, subjective-synthesized, and pure-objective). A total of 582 jury listening tests were conducted on the recorded natural sounds, human sounds and mechanical sounds. The statistical analysis results validated the reliability of constructing E-, P-, A-, and EPA-scores in the EPA model. The holistic models (O1-O6-In) demonstrated the highest goodness-of-fit among the approaches. The subjective-synthesized models (O1-O6-EPA) were designed to integrate the perceptual influences of all fundamental perceptions without the notable drop in goodness-of-fit (−6.1 % to −14.7 %) as EPA-score was a significant predictor of all negative impacts. Compared with the models relying solely on the A-weighted, equivalent continuous sound pressure level (LAeq), the pure-objective models (O1-O6-EPApre) incorporating the predictive EPA-score (EPApre) demonstrated higher adjusted R2 values (+24 % to +58 %). The results will be beneficial for future urban planning or building restoration, particularly in sustainable noise control considering environmental sound quality and noise impacts on building occupants.