Abstract
Rigid spherical microphone arrays (RSMAs) have been widely used in ambisonics sound field recording. While it is desired to combine the information captured by a grid of densely arranged RSMAs for expanding the area of accurate reconstruction, or sweet-spots, this is not trivial due to inter-array interference. Here we propose multiple scattering ambisonics, a method for three-dimensional ambisonics sound field recording using multiple acoustically interacting RSMAs. Numerical experiments demonstrate the sweet-spot expansion realized by the proposed method. The proposed method can be used with existing RSMAs as building blocks and opens possibilities including higher degrees-of-freedom spatial audio.
Introduction
In spatial audio, Ambisonics serves as the standard for capturing and reproducing 3D sound fields. This is typically achieved using a Rigid Spherical Microphone Array (RSMA)—a sphere embedded with numerous microphone capsules. While RSMAs are highly effective for 3-Degrees-of-Freedom (3DoF) experiences, such as head rotation in VR, they struggle with "translation," or moving the listener's head through space.
The primary bottleneck is the sweet spot: the physical region where the reconstructed sound field remains accurate. As sound frequency increases, this sweet spot shrinks significantly. To allow a listener to walk through a virtual environment, a single RSMA is insufficient. However, placing multiple spheres in a room introduces "inter-array interference." Because RSMAs are acoustically "hard" objects, they reflect and scatter sound waves, creating complex interference patterns that traditional algorithms cannot accurately decode.
Proposed Method: Multiple Scattering Ambisonics (MS-HOA)
The researchers propose Multiple Scattering Higher-Order Ambisonics (MS-HOA) to overcome these physical limitations. Rather than ignoring inter-array interference, the method treats the entire grid of microphone arrays as a single, complex physical system.
- The Forward Problem: The system models how an incident sound wave interacts with a grid of spheres using analytical wave physics. By employing "translation operators," the model describes how sound moves between the coordinate systems of different spheres, specifically accounting for multiple scattering—the phenomenon where sound bounces between spheres repeatedly.
- The Inverse Problem (Encoding): By mathematically modeling the interference, the researchers developed an encoding matrix using regularized least squares. This matrix processes raw signals from every microphone in the grid and "subtracts" the scattering effects to estimate the original, clean incident sound field.
- Truncation Strategy: To balance computational complexity with accuracy, the method utilizes specific truncation degrees for the incoming sound field and the scattering mathematics.
Experimental Validation
The authors conducted numerical simulations to test two hardware configurations: a Linear Grid of 6 RSMAs and a Cartesian Grid of 9 RSMAs (3x3). Each simulated sphere utilized a high-density "Spherical Fibonacci grid" of 252 microphone capsules. A monopole sound source was placed 10 meters away to simulate the incident field.
Key Results
The primary metric for success was the Sweet-Spot Area (SSA), defined as the region where the Signal-to-Distortion Ratio (SDR) exceeds 30 dB.
- Massive Expansion: For a challenging 4kHz high-frequency sound, a single sphere (HOA) produced a sweet spot of only 0.080 m². The 6-sphere MS-HOA linear grid expanded this to 0.606 m²—a 7.5x increase.
- Planar Freedom: The 9-sphere Cartesian grid achieved a wide planar sweet spot of 0.838 m², allowing for significant lateral and longitudinal movement while maintaining high audio fidelity.
- High-Frequency Stability: Unlike traditional Ambisonics, where sweet spots collapse at high frequencies, MS-HOA maintained a usable area even at 4kHz, which is vital for realistic sound localization.
Conclusion and Future Work
MS-HOA demonstrates that inter-array interference is not merely noise but a predictable physical phenomenon that can be leveraged to improve audio reconstruction. This modular approach allows existing RSMAs to be used as building blocks for large-scale spatial audio capture.
Future research will focus on reducing the hardware and computational costs. Currently, the requirement for hundreds of microphone channels and high processing power for real-time scattering equations presents a barrier to consumer adoption. Potential solutions include "thinning" the arrays (using fewer microphones per sphere) and optimizing the underlying algorithms for greater efficiency.
Summary updated