Romigh Griffin D, Simpson Brian D
Air Force Research Laboratory Dayton, OH, USA.
Front Neurosci. 2014 Dec 1;8:370. doi: 10.3389/fnins.2014.00370. eCollection 2014.
It is widely acknowledged that individualized head-related transfer function (HRTF) measurements are needed to adequately capture all of the 3D spatial hearing cues. However, many perceptual studies have shown that localization accuracy in the lateral dimension is only minimally decreased by the use of non-individualized head-related transfer functions. This evidence supports the idea that the individualized components of an HRTF could be isolated from those that are more general in nature. In the present study we decomposed the HRTF at each location into average, lateral and intraconic spectral components, along with an ITD in an effort to isolate the sound localization cues that are responsible for the inter-individual differences in localization performance. HRTFs for a given listener were then reconstructed systematically with components that were both individualized and non-individualized in nature, and the effect of each modification was analyzed via a virtual localization test where brief 250 ms noise bursts were rendered with the modified HRTFs. Results indicate that the cues important for individualization of HRTFs are contained almost exclusively in the intraconic portion of the HRTF spectra and localization is only minimally affected by introducing non-individualized cues into the other HRTF components. These results provide new insights into what specific inter-individual differences in head-related acoustical features are most relevant to sound localization, and provide a framework for how future human-machine interfaces might be more effectively generalized and/or individualized.
人们普遍认为，需要进行个性化的头部相关传递函数（HRTF）测量，以充分捕捉所有三维空间听觉线索。然而，许多感知研究表明，使用非个性化的头部相关传递函数只会使横向维度的定位精度略有下降。这一证据支持了这样一种观点，即HRTF的个性化成分可以与那些本质上更具普遍性的成分区分开来。在本研究中，我们将每个位置的HRTF分解为平均、横向和圆锥内频谱成分，以及一个耳间时间差（ITD），以努力分离出导致个体定位性能差异的声音定位线索。然后，使用本质上既个性化又非个性化的成分系统地重建给定听众的HRTF，并通过虚拟定位测试分析每次修改的效果，在该测试中，用修改后的HRTF呈现250毫秒的短噪声脉冲。结果表明，对HRTF个性化至关重要的线索几乎完全包含在HRTF频谱的圆锥内部分，并且将非个性化线索引入其他HRTF成分对定位的影响很小。这些结果为与声音定位最相关的头部相关声学特征的具体个体差异提供了新的见解，并为未来人机界面如何更有效地进行通用化和/或个性化提供了一个框架。
