Filtered Noise Shaping for Time Domain Room Impulse Response Estimation From Reverberant Speech

Abstract

Deep learning approaches have emerged that aim to transform an audio signal so that it sounds as if it was recorded in the same room as a reference recording, with applications both in audio postproduction and augmented reality. In this work, we propose FiNS, a Filtered Noise Shaping network that directly estimates the time domain room impulse response (RIR) from reverberant speech. Our domain-inspired architecture features a time domain encoder and a filtered noise shaping decoder that models the RIR as a summation of decaying filtered noise signals, along with direct sound and early reflection components. Previous methods for acoustic matching utilize either large models to transform audio to match the target room or predict parameters for algorithmic reverberators. Instead, blind estimation of the RIR enables efficient and realistic transformation with a single convolution. An evaluation demonstrates our model not only synthesizes RIRs that match parameters of the target room, such as the T60 and DRR, but also more accurately reproduces perceptual characteristics of the target room, as shown in a listening test when compared to deep learning baselines.

Blind Directional Room Impulse Response Parameterization from Relative Transfer Functions

“…To solve the estimation problem of transferplausible rendering, we propose a method for blind DRIR estimation based on beamforming and deconvolution, or rather adaptive system identification. This straightforward, interpretable approach is an alternative to algorithms for transferring acoustics from one signal to another that are entirely based on deep neural networks [2,3].…”