Localisation of Vertical Auditory Phantom Image with Band-limited Reductions of Vertical Interchannel Crosstalk

Direct sound that is captured by the upper layer of a three-dimensional (3D) microphone array is typically regarded as vertical interchannel crosstalk (VIC), since it tends to produce an undesired effect of the sound source image being elevated from the ear-level loudspeaker layer position (0°) in reproduction. The present study examined the effectiveness of band-limited VIC attenuation methods on preventing the vertical image shift problem. In a subjective experiment, five natural sound sources were presented as vertically-oriented phantom images while using two stereophonic loudspeaker pairs elevated at 0° and 30° in front of the listener. The upper layer signal (i.e., VIC) was attenuated in various octave-band-dependent conditions that were based on vertical localisation thresholds obtained from previous studies. The results showed that it was possible to achieve the goal of panning the phantom image at the same height as the image produced by the main loudspeaker layer by attenuating only a single octave band with the centre frequency of 4 kHz or 8 kHz or multiple bands at 1 kHz and above. This has a useful practical implication in 3D sound recording and mixing where a vertically oriented phantom image is rendered.