Higher-order processing of spatial impulse responses

This article pertains to parametric rendering of microphone array impulse responses, such that the spatial characteristics of a captured space may be imposed onto a monophonic input signal and reproduced over an array
of loudspeakers. Parametric methods operate by analysing a set of spatial parameters, dividing the response into components based on an assumed sound-field model, and rendering the components to the loudspeaker
array using techniques informed by the analysis. For a direct/diffuse model, the sound is divided into nondiffuse and diffuse components, which are reproduced using directional and surrounding reproduction methods,
respectively. In many cases, the input is first divided into frequency bands for the analysis and reproduction. In this article, a method capable of accommodating higher-order spherical harmonic input is proposed, which, based upon initial testing, appears to overcome some of the limitations exhibited by existing methods. The proposed approach operates by partitioning the sound-field into multiple directionally biased sectors, which are then analysed independently. The non-diffuse components are reproduced individually for each sector using amplitude-panning, whereas the diffuse components are encoded back into the spherical harmonic domain, and subsequently reproduced via linear decoding, followed by decorrelation. An open-source implementation of the proposed method is also described.