Hofman P M, Van Opstal A J
University of Nijmegen, Department of Medical Physics and Biophysics, The Netherlands.
J Acoust Soc Am. 1998 May;103(5 Pt 1):2634-48. doi: 10.1121/1.422784.
This paper describes the effect of spectro-temporal factors on human sound localization performance in two dimensions (2D). Subjects responded with saccadic eye movements to acoustic stimuli presented in the frontal hemisphere. Both the horizontal (azimuth) and vertical (elevation) stimulus location were varied randomly. Three types of stimuli were used, having different spectro-temporal patterns, but identically shaped broadband averaged power spectra: noise bursts, frequency-modulated tones, and trains of short noise bursts. In all subjects, the elevation components of the saccadic responses varied systematically with the different temporal parameters, whereas the azimuth response components remained equally accurate for all stimulus conditions. The data show that the auditory system does not calculate a final elevation estimate from a long-term (order 100 ms) integration of sensory input. Instead, the results suggest that the auditory system may apply a "multiple-look" strategy in which the final estimate is calculated from consecutive short-term (order few ms) estimates. These findings are incorporated in a conceptual model that accounts for the data and proposes a scheme for the temporal processing of spectral sensory information into a dynamic estimate of sound elevation.
本文描述了频谱-时间因素对人类二维（2D）声音定位性能的影响。受试者对呈现于额叶半球的声学刺激做出眼跳运动反应。水平（方位角）和垂直（仰角）刺激位置均随机变化。使用了三种类型的刺激，它们具有不同的频谱-时间模式，但宽带平均功率谱形状相同：噪声脉冲、调频音和短噪声脉冲序列。在所有受试者中，眼跳反应的仰角成分随不同的时间参数而系统变化，而方位角反应成分在所有刺激条件下均保持同等精度。数据表明，听觉系统并非通过对感官输入进行长期（100毫秒量级）整合来计算最终的仰角估计值。相反，结果表明听觉系统可能采用“多次观察”策略，即最终估计值由连续的短期（几毫秒量级）估计值计算得出。这些发现被纳入一个概念模型，该模型解释了数据，并提出了一种将频谱感官信息进行时间处理以形成声音仰角动态估计的方案。
