Band-Count Dense Modal Estimation with Fixed-Frequency Differentiable Resonator Refinement

2026-08-01

AI Summary

This work addresses the challenge of accurately estimating modal parameters—frequency, decay rate, gain, and total count—in dense plate reverberation impulse responses, where conventional sparse peak detection severely underestimates modal density due to weak and overlapping modes. To overcome this, the authors propose a decoupled strategy: first, an ExtraTrees regressor trained on simulated data predicts the number of modes within four frequency bands to guide the generation of a dense frequency grid; then, with frequencies fixed, a differentiable all-pole resonator model finely optimizes decay rates and gains. By separating modal density estimation from continuous parameter fitting, the approach substantially mitigates missed detections. On two synthetic test sets, it reduces local error in the challenging regime by approximately 66% compared to the official peak-detection baseline, primarily owing to significantly improved accuracy in modal count estimation.

Abstract

Task B of the 1st DAFx Parameter Estimation Challenge requires estimating the frequencies, decay rates, gains, and number of modes in a dense plate-reverb impulse response. Weak and overlapping modes make sparse peak detection prone to severe undercounting. We train an ExtraTrees regressor on simulator-generated data to predict mode counts in four frequency bands. These counts define dense frequency grids, after which a differentiable all-pole resonator model refines decay and gain while keeping frequency fixed. On two separate synthetic validation sets, the system reduces a local challenge-style error by about 66% relative to the official default peak-picking baseline. The improvement is mainly associated with lower mode-count mismatch, while decay and gain remain the largest error sources. These findings support separating modal-density estimation from continuous parameter fitting.

Problem

• modal estimation
• dense reverb
• mode counting
• overlapping modes
• parameter estimation

Innovation

• modal estimation
• differentiable resonator
• mode counting
• dense reverb
• parameter decoupling