# HRIR-Former: Grid-Free Time-Domain Reconstruction of Head-Related Impulse Responses with a Spatially Encoded Transformer

arXiv: https://arxiv.org/abs/2603.27998 · HTML: https://mlsanity.com/paper/2603.27998/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 48 references that are in the map.
- First posted on arXiv 30 Mar 2026.

## Builds on (most cited first, 11 of 11)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [HRTF upsampling with a generative adversarial network using a gnomonic equiangular projection](https://mlsanity.com/paper/2306.05812/) (2023)
- [NIIRF: Neural IIR Filter Field for HRTF Upsampling and Personalization](https://mlsanity.com/paper/2402.17907/) (2024)
- [Magnitude-Corrected and Time-Aligned Interpolation of Head-Related Transfer Functions](https://mlsanity.com/paper/2303.09966/) (2023)
- [HRTF Field: Unifying Measured HRTF Magnitude Representation with Neural Fields](https://mlsanity.com/paper/2210.15196/) (2022)
- [Retrieval-Augmented Neural Field for HRTF Upsampling and Personalization](https://mlsanity.com/paper/2501.13017/) (2025)
- [Modeling of Individual HRTFs based on Spatial Principal Component Analysis](https://mlsanity.com/paper/1910.09484/) (2019)
- [A Wide Dataset of Ear Shapes and Pinna-Related Transfer Functions Generated by Random Ear Drawings](https://mlsanity.com/paper/2003.06182/) (2020)
- [Head-Related Transfer Function Individualization Using Anthropometric Features and Spatially Independent Latent Representation](https://mlsanity.com/paper/2508.16176/) (2025)
- [RIR-Former: Coordinate-Guided Transformer for Continuous Reconstruction of Room Impulse Responses](https://mlsanity.com/paper/2602.01861/) (2026)
- [Towards HRTF Personalization using Denoising Diffusion Models](https://mlsanity.com/paper/2501.02871/) (2025)

## Abstract (arXiv)

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 HRIR-Former, 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 preprocessing is unnecessary.