This paper tackles a practical problem in spatial audio: how to convert recordings from the small, irregular microphone arrays found in real devices into Ambisonics, a device-independent format used for immersive and 3D audio. Existing neural methods can work well for a specific microphone layout, but often need to be retrained when the array changes, and methods that only use microphone positions cannot describe more realistic effects such as microphone directionality or sound scattering caused by the device body. The proposed method addresses this by giving the neural network not only the recorded audio but also a detailed description of how the microphone array responds to sound coming from different directions and frequencies. Separate neural encoders process the audio and these directional characteristics, and cross-attention combines the two to create an array-independent spatial representation from which the Ambisonics signal is generated. The method is evaluated on simulated microphone arrays in both free-field conditions and a more realistic mobile-phone scenario with complex body scattering. It achieves the best overall signal reconstruction quality in both settings, outperforming conventional Ambisonics encoding and, in realistic phone-like conditions, existing neural approaches. Importantly, the model can generalize to previously unseen microphone configurations without retraining and does so with fewer than one million parameters, showing that detailed knowledge of a device’s directional acoustic behaviour can enable compact and flexible neural spatial-audio capture systems.

This patent addresses a practical limitation of machine-learning-based spatial audio capture: neural models are typically designed for a particular microphone arrangement, so using the same technology on a different phone or recording device may require generating new training data and retraining the model. The proposed approach makes the microphone configuration itself an input to the neural network. A dedicated geometry encoder processes information describing the microphone array—such as the number and positions of the microphones and, in some embodiments, their directional acoustic responses—while a separate encoder processes the recorded audio. The two sources of information are combined by the model, which predicts a set of filters that transforms the microphone recordings into a spatial audio representation such as Ambisonics. The model can be trained using many simulated sound scenes and different microphone configurations so that it learns not only how to reconstruct spatial audio, but also how that reconstruction should change when the microphone arrangement changes. The main benefit is therefore a spatial-audio capture system that can be deployed across devices with different microphone layouts with little or no retraining, reducing the engineering effort normally required to adapt a neural spatial-audio system to each new device.