AES Fall Convention 2023

Buildings will collapse one day. Even if we try to renovate them, the original acoustic texture cannot be fully restored. Therefore, archiving the room acoustics of a historic studio is important to the audio industry.
Sometimes, an impulse response (IR) is measured to archive room reverberation. However, it also contains unwanted information, i.e. frequency responses of measurement devices and background noise. In addition, the spatial characteristics of the reverberation are limited by the microphone array we use to measure the IRs. For example, if we use a 5.0ch array, we will lose the 3D properties of the reverberation.
Instead of conventional IR reverbs, we propose a new method of archiving a room reverberation, VSVerb (Virtual sound Sources’ reVerb). The VSVerb is the method to sample the spatial properties of virtual sound sources of a room and generate a 4pi scene-based reverberation without noise and without responses of measurement tools.
In the workshop, we introduce our trial to archive the room acoustics of a recording studio in Tokyo, whose name is ONKIO HAUS. The studio has two legendary recording booths, Studio 1 and Studio 2. This time, we sampled their virtual sound sources at various source and receiver positions to archive their reverberant field. A total of 14 sets of virtual sound sources were obtained from measured IRs using an A-format microphone. These virtual sound sources were translated into time responses, then 14 kinds of VSVerbs were generated for the archive resources of the studio acoustics.
Since the VSVerb is a 4pi scene-based reverb, we can apply many types of post-processing. This time, two types of post-processing, 1) moving and rotating the microphone’s position and 2) changing the microphone’s directivity, were tested to see if the archived reverbs could provide virtual recording experiences at the ONKIO HAUS.
The listening check was performed by several recording/mixing engineers of the ONKIO HAUS in their mixing room using 2ch decoded VSVerbs. Their listening impressions were quite positive, they felt as if they were recording in the real booths of Studio 1 and Studio 2. We will also report their impressions in detail at the workshop.

Appendix: About VSVerb (Virtual sound Sources’ reVerb)
VSVerb is a sound intensity oriented technique for generating a 4pi scene-based reverb by detecting virtual sound sources from four impulse responses measured by an A-format microphone.
An overview of the processing flow is given below.
( Some papers on VSVerb can be found in the AES E-Library by typing VSV in a search window.)
1) Impulse responses are measured in a target room using an A-format microphone, and they are filtered into low, mid, and high frequency bands.
2) The filtered impulse responses are converted to W, X, Y and Z of B-format signals, and Hilbert transforms are applied to them.
4) By multiplying the Hilbert transformed W by X, Y, Z, three orthogonal sound intensities, Ix, Iy and Iz, are obtained. Then time averaging operations are applied to the sound intensities.
5) Virtual sound sources, i.e. dominant reflections, are detected from the averaged sound intensities using the "Speed detection" method.
“Speed detection” is the method to detect acoustic information of real and virtual sound sources, i.e. strength, distance, arrival direction and phase (+/-), from measured sound intensities by focusing on moving speeds of sound intensities.
6) The spatial information of the obtained virtual sound sources is translated into time responses, then the 4pi scene-based reverberation is generated in low, mid, and high frequency bands.
7) Since VSVerb is a 4pi scene-based reverb, the reverb can be divided into any types of playback channel formats and has high flexibility for post-processing operations, e.g. moving and rotating the receiver’s position, changing the receiver’s directivity, changing the averaged absorption coefficient of a room, changing the room size, etc.