自然与城市声特征及自动识别

Characteristics of Natural and Urban Sounds and the Automatic Identification

从心理声学、音乐特性和1/f噪声动态三个方面阐述了自然和城市环境中各种声音类型的特征及其之间的差异。一般而言,城市声同时具有高波动强度和高响度,而自然声或具有低波动强度或低响度。基于这些不同声音类型各自的特征,此研究利用计算机算法使各类声音得到自动识别,精确度在97%以上。其结果为声景设计和进一步的研究提供了一定的理论依据及参考和一种新的更为直接的方法。

This paper studies the characteristics of various natural and urban sounds and their differences with respect to psychoaeoustics, music features, and 1/f noise dynamic. In general, urban sounds have both high loudness and fluctuation, whereas natural sounds have low value of at least one of the two indicators. In addition to this basic understanding, the sound types are automatically identified and classified, with a prediction accuracy over 97%. The research results provide a reference and a new method for soundscape design and research.