Blind Estimation of Sound Coloration in Rooms

Doctoral thesis, 2024

A common problem in sound reinforcement systems consisting of one or more
microphone-amplifier-loudspeaker channels is the sound coloration caused by
the repetitive amplification of strong frequency components in the loudspeakermicrophone
transfer function(s). For hidden systems, such as certain reverberation
enhancement systems, acoustic feedback, along with other causes for sound
coloration, risk compromising the impression of a natural sounding acoustic
environment. Therefore, in this thesis, a methodology for blind estimation of
sound coloration is developed and evaluated. Depending on the room type and
various other assumptions, the damping distribution of a room will follow a
specific ”reference distribution,” i.e. any deviation from the distribution should
indicate sound coloration. Using one microphone placed in the audience area,
blind estimation of sound coloration is achieved by computing decay times of
non-harmonic components in the time-frequency domain. The results show
that the computed damping distributions agree well with the chi-square distributions
at low system gains. As the system gain increases, the distributions
are shifted toward lower damping constants, and their shapes deviate more and
more from the reference distribution, thus, giving a clear indication of sound
coloration. The suggested objective measures show that deviations from the
reference damping distribution can be detected at substantially lower system
gains compared to results of related listening tests where audible coloration
is evaluated. Thus, it is safe to conclude that the proposed methodology for
detecting and classifying sound coloration performs well for the studied cases.
Further research is needed to optimize its robustness when using different room
types, system configurations, and so on.