头相关传输函数表征了声源到耳的传输特性，在虚拟声重放及其相关领域具有广泛的应用。针对头相关传输函数的多维特征，提出了基于独立成分分析的中垂面谱特征的提取和建模。计算结果表明：中垂面的平均谱失真随着展开阶数的增加呈逐渐下降的趋势，前六阶独立成分的计权组合可有效表征中垂面谱的主要特征。进一步，采用主观辨别实验验证了模型的听觉有效性。实验结果表明，在±45°的仰角范围中，所提出的中垂面谱模型不会引起听觉感知的畸变。

Head-related transfer functions (HRTFs) reflect the transmission process from sound sources to human ears, and therefore vary with sound source position, individual, and frequency. Aiming to obtain statistically independent information encoded in HRTFs, an extracting and modeling method of median-plane spectral characteristics was proposed on the basis of independent component analyses (ICA). Calculation results indicate that the mean of median-plane spectral distortion gradually decreases with increasing expansion order, and a weighted combination of the preceding six orders components represents the main part of median-plane spectral characteristics. Moreover, a subjective discrimination experiment was carried out. Corresponding results show that no auditory degradation is perceived when the proposed ICA-based median-plane spectral model is used in the elevation range from 45° to ?45°.