Search papers, labs, and topics across Lattice.
This paper addresses the sensitivity of coordinate-conditioned neural networks for personal sound zone (PSZ) generation to localization uncertainty. They propose a neighbor-consistent training regularization that penalizes filter differences at randomly perturbed neighboring coordinates to improve robustness against tracking noise. Experiments in simulation and in-situ measurements demonstrate that this regularization significantly reduces spatial variation rates and improves worst-case neighborhood isolation, stabilizing PSZ rendering under localization uncertainty.
Even slight tracking errors can drastically degrade personal sound zone performance, but a simple neighbor-consistency regularization can dramatically improve robustness.
Coordinate-conditioned neural networks can generate head-tracked personal sound zone (PSZ) loudspeaker filters in real time, but they are sensitive to localization uncertainty. Small fluctuations in estimated listener coordinates, caused by optical distortion, temporary occlusions, or tracking jitter, may produce large filter changes even when listeners are physically stationary. This paper proposes neighbor-consistent neural filters that regularize the coordinate-to-filter mapping by penalizing filter differences at randomly perturbed neighboring coordinates during training. To evaluate robustness against tracking noise, we introduce a decoupled protocol that fixes the acoustic transfer functions at a physical anchor while perturbing only the coordinate inputs used for filter generation. Isolation quality and local stability are evaluated using neighborhood median and lower-tail statistics of inter-zone and inter-program isolation, together with spatial variation rates that quantify metric sensitivity within a coordinate neighborhood. In simulation with a split-band woofer-tweeter system and 25 randomly sampled anchor positions, neighbor consistency reduces the root-mean-square (RMS) variation rate by up to 55.9% in the woofer band and 30.3% in the tweeter band while largely preserving isolation quality and improving lower-tail robustness. In in-situ measurements using a 24-driver array and two stationary head-and-torso simulators, the proposed regularization improves worst-case neighborhood isolation by up to 16.9% and reduces spatial variation rates by up to 61.8%. These results demonstrate that neighbor-consistency regularization effectively stabilizes PSZ rendering under localization uncertainty.
