Differentiable Time-Varying IIR Filtering for Real-Time Speech Denoising

TL;DR

This paper introduces TVF, a low-latency, interpretable neural model that dynamically predicts IIR filter coefficients for real-time speech denoising, bridging traditional DSP and deep learning.

Contribution

The paper presents a novel differentiable, time-varying IIR filtering approach that combines interpretability with adaptability in neural speech enhancement models.

Findings

TVF outperforms static and fully deep-learning methods in non-stationary noise conditions.

The model achieves real-time processing with 1 million parameters.

Demonstrates effective noise adaptation on the Valentini-Botinhao dataset.

Abstract

We present TVF (Time-Varying Filtering), a low-latency speech enhancement model with 1 million parameters. Combining the interpretability of Digital Signal Processing (DSP) with the adaptability of deep learning, TVF bridges the gap between traditional filtering and modern neural speech modeling. The model utilizes a lightweight neural network backbone to predict the coefficients of a differentiable 35-band IIR filter cascade in real time, allowing it to adapt dynamically to non-stationary noise. Unlike ``black-box'' deep learning approaches, TVF offers a completely interpretable processing chain, where spectral modifications are explicit and adjustable. We demonstrate the efficacy of this approach on a speech denoising task using the Valentini-Botinhao dataset and compare the results to a static DDSP approach and a fully deep-learning-based solution, showing that TVF achieves effective adaptation to changing noise conditions.