On the Usefulness of Diffusion-Based Room Impulse Response Interpolation to Microphone Array Processing

TL;DR

This paper presents a diffusion-based interpolation method for Room Impulse Responses that improves spatial audio processing and speech enhancement, validated on real-world data.

Contribution

It extends a diffusion-based inpainting framework to practical multi-microphone array tasks and demonstrates robustness with real-world Room Impulse Responses.

Findings

Improved Room Impulse Response interpolation performance.

Enhanced spatial audio processing and speech enhancement results.

Validated robustness on real-world Room Impulse Responses.

Abstract

Room Impulse Responses estimation is a fundamental problem in spatial audio processing and speech enhancement. In this paper, we build upon our previously introduced diffusion-based inpainting framework for Room Impulse Response interpolation and demonstrate its applicability to enhancing the performance of practical multi-microphone array processing tasks. Furthermore, we validate the robustness of this method in interpolating real-world Room Impulse Responses.