Estimation of sound source elevation by extracting the vertical localization cues from binaural signals

The author has proposed a parametric HRTF model for vertical sound localization. The parametric HRTF is recomposed only of the spectral peaks and notches extracted from the measured HRTF. The results of the median plane localization tests, which were carried out using the parametric HRTFs with various combinations of spectral peaks and notches, show that the pair of first and second notches (N1 and N2) above 5 kHz can be regarded as spectral cues. Then, utilizing these findings, estimation of the elevation of sound source in the upper median plane by extracting N1 and N2 frequencies from binaural input signal was carried out. The kinds of sound sources were female voice, male voice, music, white noise, and pink noise. The results show that the estimation is accurate for almost of all the elevations and of all the kind of sound sources.