Abstract
This article introduces three methods for artifact reduction in Differential Head-Related Transfer Function (DHRTF). Such artifacts may occur within virtual sound source positioning methods when spike-like spectral peaks occur in the module of the DHRTF.
These spectral features result in an undesirable whistling-like sound component that distorts both timbre and spatial perception of the virtual sound. DHRTF filters require preprocessing by an appropriate method that is able to remove or reduce the artifacts while preserving both the natural character of the timbre and the spatial perception of the positioned sound.
Preprocessing of the DHRTF set avoids the effects of the negative ILD while preserving all the necessary localization cues and natural timbre of the positioned sound. Three methods based on limiting and/or smoothing of the DHRTF magnitude by a low-pass filter were introduced and evaluated by both paired comparison listening tests and objective assessment techniques.
Even elementary preprocessing algorithms improve the quality of the positioned sound.