Developing multivariate models for predicting the levels of multi-dimensional critical perceptions due to metro noise inside buildings

Abstract

With the rapid expansion of urban metro systems, the negative emotions induced by metro noise inside buildings have caused widespread concern. Although a large number of models have been established to predict the various negative perceptions caused by road traffic or aircraft noises, there is a lack of model to evaluate the diverse negative perceptions due to metro noise. Accordingly, this study selected the multi-dimensional perceptions which could better capture the metro noise responses inside buildings as critical perceptions and developed the optimal multivariate models for predicting the levels of multi-dimensional critical perceptions. Seventy-nine subjects participated in the experiment and were exposed to a series of metro noise samples before being asked to cast their multi-dimensional perception votes and activity disturbance votes. The results showed that compared to rumble loudness, low-frequency perception, annoyance and non-acceptance, subjective loudness, depression, discomfort and dissatisfaction had superior psychometric properties (validity and internal consistency) and were determined as the critical perceptions to describe the metro noise reactions. Besides, as for the levels of multi-dimensional critical perceptions, they correlated well with A-weighted, C-weighted and Z-weighted equivalent sound pressure level (LAeq, LCeq and LZeq), among psychoacoustic measures, loudness and roughness primarily impacted them, and the sensitivity, stance and adaptability to metro noise were their main non-acoustic influencing factors. Furthermore, the levels of multi-dimensional critical perceptions could be evaluated most effectively based on LAeq, compared to the levels of depression and discomfort, the levels of subjective loudness and dissatisfaction were more sensitive to the change of LAeq (slightly higher slopes). After adding the three major non-acoustic influencing factors (the sensitivity, stance and adaptability to metro noise), the coefficient of determination (R2) values of the prediction models for the levels of multi-dimensional critical perceptions reached their highest levels, and the levels of multi-dimensional critical perceptions could be well estimated with R2 values close to or greater than 0.5.