Perceptual equalization in near-speaker panning

Bernhard Seeber and Ervin Hafter


In recent years, extensive work has been done on spatial sound
reproduction techniques involving multiple loudspeakers. While many
studies on panning focus only on directional reproduction, we ask if
panning can truly replace a loudspeaker placed between two speakers
spaced 7.5deg or 15deg apart. This requires not only that panned
direction and speaker position correspond, but also that source
width, loudness, timbre, and temporal aspects are equally reproduced.
Comparisons in an anechoic chamber between panned sources and sounds 
played from a loudspeaker at the panned location revealed small 
perceptual differences, primarily in timbre. These stemmed from the 
effects of the head. 

In order to examine the errors in panning, signals were shaped with
FIR-filters derived with three different approaches: 
(1) Compensation filters were determined from a listening experiment 
in which the level of narrow band noises was found to yield equal 
loudness with and without panning. 
(2) Level compensation was done based on panning simulations with 
KEMAR-HRTFs, or
(3) with a spherical head model. 

Listening tests showed that all approaches succeeded in reducing
perceived panning errors to near inaudibility. This talk will address
the link between the remaining perceptual error and the spectral and
temporal composition of the sounds, the sound direction, and the
panning angle. Additional listening tests showed that level roving or
the introduction of reflections rendered panning errors inaudible. 
We conclude that a simple equalization is sufficient to render panned 
sources from nearby speakers perceptually equivalent to real sources.