Middlebrooks J C
Department of Neuroscience, University of Florida, Gainesville 32610-0244.
J Acoust Soc Am. 1992 Nov;92(5):2607-24. doi: 10.1121/1.404400.
Human subjects localized brief 1/6-oct bandpassed noise bursts that were centered at 6, 8, 10, and 12 kHz. All testing was done under binaural conditions. The horizontal component of subjects' responses was accurate, comparable to that for broadband localization, but the vertical and front/back components exhibited systematic errors. Specifically, responses tended to cluster within restricted ranges that were specific for each center frequency. The directional transfer functions of the subjects' external ears were measured for 360 horizontal and vertical locations. The spectra of the sounds that were present in the subjects' ear canals, the "proximal stimulus" spectra, were computed by combining the spectra of the narrow-band sound sources with the directional transfer functions for particular stimulus locations. Subjects consistently localized sounds to regions within which the associated directional transfer function correlated most closely with the proximal stimulus spectrum. A quantitative model was constructed that successfully predicted subjects' responses based on interaural level difference and spectral cues. A test of the model, using techniques adapted from signal detection theory, indicated that subjects tend to use interaural level difference and spectral shape cues independently, limited only by a slight spatial correlation of the two cues. A testing procedure is described that provides a quantitative comparison of various predictive models of sound localization.
人类受试者对以6千赫、8千赫、10千赫和12千赫为中心的短暂1/6倍频带通带噪声突发进行定位。所有测试均在双耳条件下进行。受试者反应的水平分量是准确的，与宽带定位的情况相当，但垂直分量和前后分量表现出系统误差。具体而言，反应倾向于聚集在每个中心频率特有的受限范围内。测量了受试者外耳在360个水平和垂直位置的方向传递函数。通过将窄带声源的频谱与特定刺激位置的方向传递函数相结合，计算出受试者耳道内存在的声音频谱，即“近端刺激”频谱。受试者始终将声音定位到与相关方向传递函数与近端刺激频谱相关性最密切的区域。构建了一个定量模型，该模型基于双耳声级差和频谱线索成功预测了受试者的反应。使用信号检测理论改编的技术对该模型进行的测试表明，受试者倾向于独立使用双耳声级差和频谱形状线索，仅受这两种线索的轻微空间相关性限制。描述了一种测试程序，该程序可对各种声音定位预测模型进行定量比较。
