Improvements in calculating the loudness of time varying sounds
Improvements in calculating the loudness of time varying sounds
复制标题
计算时变声音响度的改进
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
R. Sottek
中科院分区：
文献类型：
--
作者：
R. Sottek
最近，为修订ISO 532：1975第2节(方法B)(2)，新的ISO标准ISO 532-1(1)被提出用于任意声音的响度。它基于DIN 45631/A1：2010(3)，其中包括作为特例的针对静止声音的广泛使用的标准DIN 45631：1991(4)。ISO 532-1通过严格定义响度计算的完整程序，从时间信号的波形开始，以特定和总响度与时间函数的关系结束，消除了现有标准的不确定性。完整程序的严格定义是向计算响度结果的可比性迈进的一步，为了连续性，完全符合DIN 45631/A1：2010。然而，尽管该算法的结果与许多听力测试的结果相一致，但仍有一些现象没有被该方法覆盖。例如，扫描信号的计算响度显示波动，而感知的响度则不显示。这是由于实现了基于固定的、连续的第三倍频程滤波器的滤波器组。作为一种可能的解决方案，提出了一种基于听觉模型(Sottek)的响度计算方法，该方法使用高度重叠的非对称滤波器(5)的听觉充足的滤波器组。此外，根据多次听力测试的结果，在该模型中重新考虑了比响度和声压之间的非线性。
Recently, a new ISO standard for loudness of arbitrary sounds ISO 532-1 (1) was proposed for the revision of ISO 532:1975 section 2 (method B) (2). It is based on DIN 45631/A1:2010 (3), which includes the widely used standard DIN 45631:1991 (4) for stationary sounds as a special case. ISO 532-1 eliminates uncertainties of existing standards by strictly defining the complete procedure of loudness calculation starting with the waveform of the time signal and ending with specific and total loudness vs. time functions. The strict definition of the complete procedure is a step forward to comparability of calculated loudness results, and fully conforms to DIN 45631/A1:2010 for the sake of continuity. However, although the results of this algorithm are in accordance with the results of many listening tests, there are still phenomena that are not covered by this method. For example, the calculated loudness of sweep signals shows fluctuations, whereas the perceived loudness does not. This is due to the implemented filter bank based on fixed, contiguous third-octave filters. As a possible solution a loudness calculation method is presented that is based on a hearing model (Sottek) using an aurally adequate filter bank of highly-overlapping asymmetric filters (5). In addition, the nonlinearity between specific loudness and sound pressure has been reconsidered in this model according to results of many listening tests (6).