Synthesis of Spatially Extended Virtual Sources with Time-Frequency Decomposition of Mono Signals

Synthesis of volumetric virtual sources is a useful technique for auditory displays and virtual worlds. This task can be simplified into synthesis of perceived spatial extent. Previous research in virtual-world Directional Audio Coding has shown that spatial extent can be synthesized with monophonic sources by applying a time-frequency-space decomposition, i.e., randomly distributing time-frequency bins of the source signal. However, although this technique often achieved perception of spatial extent, it was not guaranteed and the timbre could degrade. In this article this technique is revisited in detail and the effect of different parameters is examined to ultimately achieve optimal quality and perception in all situations. The results of a series of informal and formal experiments are presented here, and they suggest that the revised method is viable in many cases. There is some dependency on the signal content that requires proper tuning of parameters. Furthermore, it is shown that different distribution widths can be produced with the method as well. From a psychoacoustical perspective, it is interesting that distributed narrow frequency bands form a spatially extended auditory event with no apparent directional focus.