NLDSI-BWE: Non Linear Dynamical Systems-Inspired Multi Resolution Discriminators for Speech Bandwidth Extension

2025/10/01 by Tamiti, Tarikul Islam, Barua, Anomadarshi
#FOS: Computer and information sciences #Sound (cs.SD)

paper · doi:10.48550/arxiv.2510.01109

Abstract

In this paper, we design two nonlinear dynamical systems-inspired discriminators -- the Multi-Scale Recurrence Discriminator (MSRD) and the Multi-Resolution Lyapunov Discriminator (MRLD) -- to explicitly model the inherent deterministic chaos of speech. MSRD is designed based on Recurrence representations to capture self-similarity dynamics. MRLD is designed based on Lyapunov exponents to capture nonlinear fluctuations and sensitivity to initial conditions. Through extensive design optimization and the use of depthwise-separable convolutions in the discriminators, our framework surpasses prior AP-BWE model with a 44x reduction in the discriminator parameter count (∼ 22M vs ∼ 0.48M). To the best of our knowledge, for the first time, this paper demonstrates how BWE can be supervised by the subtle non-linear chaotic physics of voiced sound production to achieve a significant reduction in the discriminator size.

Citations

• A High-Fidelity Speech Super Resolution Network using a Complex Global Attention Module with Spectro-Temporal Loss
• DSP-informed bandwidth extension using locally-conditioned excitation and linear time-varying filter subnetworks
• Multi-Stage Speech Bandwidth Extension with Flexible Sampling Rate Control
• Towards High-Quality and Efficient Speech Bandwidth Extension with Parallel Amplitude and Phase Prediction
• Time-domain Speech Enhancement Assisted by Multi-resolution Frequency Encoder and Decoder
• AERO: Audio Super Resolution in the Spectral Domain
• A Multi-Scale Time-Frequency Spectrogram Discriminator for GAN-based Non-Autoregressive TTS
• A ConvNet for the 2020s
• A ConvNet for the 2020s
• UnivNet: A Neural Vocoder with Multi-Resolution Spectrogram Discriminators for High-Fidelity Waveform Generation
• HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis
• Denoising Diffusion Probabilistic Models
• Audio Super Resolution using Neural Networks
• Determining Lyapunov exponents from a time series

Related

• Nonlinear Framework for Speech Bandwidth Extension
• CS-ETS: Chaos-Inspired Samba-Based EMG-To-Speech Synthesis with Nonlinear Chaotic Losses
• Physics-inspired Neuroacoustic Computing Based on Tunable Nonlinear Multiple-scattering
• On chaotic nature of speech signals
• Learnable Nonlinear Compression for Robust Speaker Verification