A model of saliency-based auditory attention to environmental sound
A model of saliency-based auditory attention to environmental sound
复制标题
基于显着性的环境声音听觉注意模型
DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
D. Botteldooren
中科院分区：
文献类型：
--
作者：
B. D. Coensel;D. Botteldooren
受人类听觉系统结构的启发，提出了一种对环境声音的听觉注意的计算模型。该模型模拟听众如何根据自下而上和自上而下的线索随时间在不同的听觉流之间切换注意力。自下而上的线索由每个流的时间相关显着性确定。后者是根据听觉显着性图计算的，该显着性图对声音的强度和频谱和时间不规则性的数量以及不同流的二进制频谱-时间掩模进行编码。自上而下的线索是由对特定听觉流的意志关注程度决定的。一种竞争性的赢者通吃机制，平衡每个流的自下而上和自上而下的注意力，决定选择哪个流进入工作记忆。因此，该模型能够界定关注特定流的时间段。尽管主要思想可以应用于所有类型的声音，但该模型的实施尤其针对环境声音。作为说明，该模型用于重现关于交通噪声感知的详细现场实验的结果。最后，展示了该模型如何成为户外声景设计中补充可听性的宝贵工具。
A computational model of auditory attention to environmental sound, inspired by the structure of the human auditory system, is presented. The model simulates how listeners switch their attention over time between different auditory streams, based on bottom-up and top-down cues. The bottom-up cues are determined by the time-dependent saliency of each stream. The latter is calculated on the basis of an auditory saliency map, which encodes the intensity and the amount of spectral and temporal irregularities of the sound, and binary spectro-temporal masks for the different streams. The top-down cues are determined by the amount of volitional focusing on particular auditory streams. A competitive winner-takes-all mechanism, which balances bottom-up and top-down attention for each stream, determines which stream is selected for entry into the working memory. Consequently, the model is able to delimit the time periods during which particular streams are paid attention to. Although the main ideas could be applied to all types of sound, the implementation of the model was targeted at environmental sound in particular. As an illustration, the model is used to reproduce results from a detailed field experiment on the perception of transportation noise. Finally, it is shown how this model could be a valuable tool, complementing auralization, in the design of outdoor soundscapes.