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°.