Complexity Reduction of Virtual Reverberation Filtering Based on Index-Based Convolution for Resource-Constrained Devices

Abstract

Virtual reverberation effects are a vital part of virtual audio reality. Reverberation effects can be directly applied by implementing a convolution process between the input audio and a reverberation filter response that characterizes a virtual space. In order to apply reverberation effects, however, additional or dedicated processors are required for practical implementation due to the excessively long impulse response of the reverberation filter. In this paper, we propose a fast method for applying virtual reverberation effects based on a reverberation filter approximation and an index-based convolution process. Throughout exhaustive experiments, we attempt to optimize the proposed method in terms of satisfaction of the reverberation effect and its computational requirements. We then implement three different types of virtual reverberation functions in a resource-constrained digital imaging device. It is shown that the virtual reverberation effects implemented by the proposed approach are able to operate in real-time with less than 5ms latency, with an over 80% overall satisfaction score in the subjective preference test.