Auditory adapted exponential transfer function smoothing (AAS)

Typically, the magnitude of a reflective acoustical system´s transfer function shows strong variation over frequency. This variation may cause technical problems when the transfer function is employed for auralization purposes, since large linear amplification is required to play back signals convolved with the corresponding impulse response. For that reason, it is common practice to apply spectral smoothing during the auralization system design process. The goal of the smoothing procedure is to reduce spectral fluctuation while not diminishing the amount of relevant auditory information in the signal. Within this contribution, the system-theoretical foundation of a smoothing procedure referred to as auditory adapted exponential smoothing (AAS) is derived, which is adapted to the spectral properties of the human hearing system by implementing frequency-dependent smoothing bandwidths. Possibly occurring time domain effects are identified and discussed, along with two approaches to cope with the resulting shortcomings.