Using ANN to study sound preference evaluation in urban open spaces

Abstract

In soundscape research, subjective preference evaluation of a sound is crucial. Based on a series of field studies and laboratory experiments, influence of sound category and psychoacoustic parameters on sound preference evaluation is examined. It has been found that sound category and loudness and sharpness are important. Regarding a previous study, age and education level are also important to influence sound preference evaluation. In order to understand user’s preference in terms of sound at a design stage, prediction of sound preference evaluation is essential. As sound preference evaluation is complicated and influenced by various factors linearly and non-linearly, artificial neural network (ANN) has been explored to make predictions of sound preference evaluation. A number of developed ANN models have been demonstrated, and it has been found that the models including input factors of sound category, loudness and sharpness produce better predictions than others. The best prediction model is the one that is based on an individual case study site. Based on the best prediction model, a mapping tool for sound preference evaluation has been developed and its usefulness for aiding landscape architects and urban designers has been demonstrated.