Comparative Study of Optimal Design Strategies of Reverberators

ABSTRACT

Artificial reverberator is a key element in spatial audio reproduction. This paper compares various optimal design strategies of artificial reverberators for room response simulation. From the comparison, it is hoped that guidelines for design of natural-sounding reverberators can be found. Infinite impulse response (IIR) filters such as allpass filters or comb filters are used as building blocks for the design. Early reflections of room responses are modeled by three different approaches: the traditional image method, the exponential truncation method, and network approximation method. On the other hand, comb/nested allpass filter networks are exploited for modeling late reverberations, with parameters predetermined by a genetic algorithm (GA)-based procedure. Up/down-sampling and spectral band replication (SBR) techniques are also employed to save memory storage and recover the lost high frequency portion of the reverberation. Subjective listening tests were conducted to compare the proposed artificial reverberator modules. The test results revealed that the reverberator in which early reflections are modeled by network approximation and late reverberations are modeled by comb/nested allpass filter network is superior in delivering natural-sounding reverberation among all methods.