BiFormer3D: Grid-Free Time-Domain Reconstruction of Head-Related Impulse Responses with a Spatially Encoded Transformer
This addresses the costly need for dense per-listener measurements in binaural rendering, offering a more efficient solution for audio applications.
The paper tackles the problem of reconstructing individualized head-related impulse responses (HRIRs) at arbitrary directions from sparse measurements, proposing BiFormer3D, which improves normalized mean squared error, cosine distance, and ITD/ILD errors over prior methods.
Individualized head-related impulse responses (HRIRs) enable binaural rendering, but dense per-listener measurements are costly. We address HRIR spatial up-sampling from sparse per-listener measurements: given a few measured HRIRs for a listener, predict HRIRs at unmeasured target directions. Prior learning methods often work in the frequency domain, rely on minimum-phase assumptions or separate timing models, and use a fixed direction grid, which can degrade temporal fidelity and spatial continuity. We propose BiFormer3D, a time-domain, grid-free binaural Transformer for reconstructing HRIRs at arbitrary directions from sparse inputs. It uses sinusoidal spatial features, a Conv1D refinement module, and auxiliary interaural time difference (ITD) and interaural level difference (ILD) heads. On SONICOM, it improves normalized mean squared error (NMSE), cosine distance, and ITD/ILD errors over prior methods; ablations validate modules and show minimum-phase pre-processing is unnecessary.