基于人工神经网络技术的商业步行街声景预测研究

商业步行街是我国大中城市重点投资建设的项目之一，同时也是重要的城市景观。然而商业步行街人数众多，噪声污染大，声环境复杂，使得人们的听觉感受较低，从而严重影响了商业步行街的使用性与观赏性。因此，如何利用绿色声学和可持续发展理念改善和预测商业步行街声环境，成为一个亟待解决的重要科学问题。本研究首先以客观测量和主观调查的方式为基础，建立商业步行街声景的数据库。其次，通过分析声音、使用者、空间和环境等因子，确定商业步行街声景的影响因素。再次，构建基于人工神经网络的商业步行街声景预测模型，并在多参数比较和神经网络模型对比分析的基础上，确定最优的预测模型。最后，本研究将根据商业步行街的特点，提出多种有针对性的分类预测模型，并通过提炼出声景的主要影响因素，结合预测结果，编制商业步行街声景预测指南。本研究期望通过人工神经网络预测模型和预测指南为商业步行街的立项决策、优化设计、项目验收和大众监督提供依据。

Commercial pedestrian streets are not only important projects but also indispensable landscapes in present cities.However, because of noise pollution as well as large number of persons, the evaluation of sound environment in commercial pedestrian streets is low, therefore, how to improve the environmental construction using the concept of green acoustics and of sustainable development becomes an important scientific problem. In this research, firstly, a soundscape database which combines characteristics of sounds, users, environment and spaces will be established based on measurements and questionnaire surveys. Secondly, the factors which influenced the soundscape in commercial pedestrian streets will be analyzed by using different methods such as statistics, building acoustics and environmental psychology. Thirdly, the optimized prediction models based on artificial neural networks (ANN) for soundscape in commercial pedestrian streets will be given by analyzing the different parameters and networks. Finally, some group prediction models will be given by considering different types of commercial pedestrian streets, while a prediction handbook for soundscape in commercial pedestrian streets will be also given based on main factors of soundscape and prediction models. From this research, the prediction models and the prediction handbook will be used for planning and designing the following commercial pedestrian streets, and for the public supervision.