Efficient Optimization of Feedback Delay Networks for Smooth Reverberation

2024-02-17

AI Summary

Spectral coloration (e.g., metallic ringing) in artificial reverberation severely degrades audio quality—particularly under hardware-constrained conditions with few delay lines (e.g., only four). This paper proposes an end-to-end differentiable feedback delay network (FDN) optimization framework that jointly learns a sparse feedback matrix, input/output gains, and dynamic decay filters. Methodologically, we couple spectral flatness loss with L₁/L₂ regularization to concentrate modal excitation and control reverberation density; we further introduce, for the first time, a learnable scattering structure to enhance impulse response (IR) naturalness. Objective and subjective evaluations demonstrate significant suppression of late-time coloration, reduced computational overhead versus baseline methods, and generation of high-fidelity, perceptually natural synthetic IRs across diverse acoustic scenarios.

Technology Category

• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Search and Retrieval-Augmented AI: Assisted, interactive, and conversational search
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

A common bane of artificial reverberation algorithms is spectral coloration, typically manifesting as metallic ringing, leading to a degradation in the perceived sound quality. This paper presents an optimization framework where a differentiable feedback delay network is used to learn a set of parameters to reduce coloration iteratively. The parameters under optimization include the feedback matrix, as well as the input and output gains. The optimization objective is twofold: to maximize spectral flatness through a spectral loss while maintaining temporal density by penalizing sparseness in the parameter values. A favorable narrower distribution of modal excitation is achieved while maintaining the desired impulse response density. In a subjective assessment, the new method proves effective in reducing perceptual coloration of late reverberation. The proposed method achieves computational savings compared to the baseline while preserving its performance. The effectiveness of this work is demonstrated through two application scenarios where natural-sounding synthetic impulse responses are obtained via the introduction of attenuation filters and an optimizable scattering feedback matrix.

Problem

• Reduces spectral coloration in artificial reverberation algorithms
• Optimizes parameters for spectral flatness and temporal density
• Achieves smooth-sounding synthetic room impulse responses

Innovation

• Optimizes tiny feedback delay networks
• Reduces spectral coloration via optimization
• Uses attenuation filters and feedback matrix