Download A Dual-Stream Framework Combining Audio Spectrogram Transformer and Dynamic Mode Decomposition for Plate Modal Parameter Estimation ★ Plate reverberation is characterized by a dense distribution of resonant modes, which makes the estimation of modal parameters from observed responses a challenging inverse problem. To address this problem, we propose a physics guided dual stream framework that integrates an Audio Spectrogram Transformer (AST) with Dynamic Mode Decomposition (DMD). The AST branch models the global temporal and spectral structure of the impulse response, whereas the DMD branch extracts local descriptors associated with modal dynamics. The resulting representations are fused and processed by convolutional prediction heads to jointly estimate mode presence and the corresponding modal parameters. Experiments on the official validation set of Task B in the DAFx Challenge show that the proposed method reduces the overall relative error from 1.976 for the official baseline to 0.867. These results demonstrate that integrating local dynamic information derived from physical modeling with global transformer based representations substantially improves plate modal parameter estimation.
Download A Comparative Study of Kolmogorov-Arnold Networks and Multi-Layer Perceptrons for Virtual Analog Modeling in Wave Digital Filters The design of Virtual Analog (VA) algorithms has traditionally been divided between white-box (physics-based) and black-box (data-driven) approaches. Recent work has shown that hybrid methods, combining physical modeling with neural networks, can effectively leverage the strengths of both paradigms. In particular, Wave Digital Filters (WDFs) can be coupled with Multi-Layer Perceptrons (MLPs) to model circuits with multiple nonlinearities in a fully explicit manner. In this paper, we present a comparative study investigating the use of Kolmogorov-Arnold Networks (KANs) for VA modeling within the WDF framework. Unlike MLPs, KANs shift the learning paradigm by parameterizing activation functions instead of relying exclusively on learned weight matrices, potentially enabling more compact representations. Results show that, for our case study, KANs achieve accuracy comparable to MLPs while requiring approximately 70% fewer parameters at the cost of increased computational complexity. These findings suggest that KANs may represent a promising alternative in scenarios where memory footprint is a primary constraint, such as embedded audio applications, or when target models feature numerous nonlinear elements.
Download Performance-Oriented Wave Digital Circuit Emulation Wave Digital Filters are a circuit-modeling paradigm well-suited for reusable software implementation, but existing software implementations often incur significant overhead due to run-time abstractions and data layout constraints. This paper presents a performance-oriented toolchain for implementing Wave Digital circuit models based on static code generation. The toolchain consists of a declarative circuit description language, a compiler that generates circuit simulation code with minimal persistent state and no run-time abstraction, and a minimal runtime library implementing specialized circuit components as Wave Digital Filters. Performance measurements across several test circuits demonstrate that the generated models consistently outperform existing implementations, and achieve near-ideal performance relative to a theoretical execution bound.
Download Stability Analysis of Time-Varying Virtual Analog Filters Time-varying virtual analog filters used in digital audio effects and synthesizers are often implemented by discretizing continuous-time state-space systems using trapezoidal integration. When filter parameters such as cutoff frequency or resonance vary over time, as is the norm in musical applications, proving BIBO stability of the resulting time-varying discrete-time system becomes nontrivial. In this paper, we review the technique of common quadratic Lyapunov functions (CQLFs) from the control systems literature and show how a continuous-time CQLF is preserved through discretization. This allows us to prove stability of some time-varying virtual analog filters by working in the often simpler continuous-time domain. We apply this framework to several filters of musical interest, providing new proofs of stability for the state variable filter and the Sallen-Key filter, and new bounds on the stable time-varying parameter range for the Moog ladder and diode ladder filters.
Download Quantifying Nonlinear Behavior in Digital Moog Ladder Filters: Cross-Implementation Comparison and Common-Core Ablation Digital Moog ladder filters are often compared by their linear frequency response, even though musically important differences emerge under nonlinear operation. This paper presents an open, reproducible SPICE-referenced evaluation framework that combines two components: a cross-implementation comparison of five digital ladder-filter implementations against the same SPICE reference, and a controlled common-core ablation. The cross-implementation comparison shows that close linear agreement can mask substantial nonlinear divergence, while the ablation shows that ladder nonlinear behavior depends not only on saturator choice and nonlinearity placement, but also on the non-uniform contribution of different ladder stages: in partial ablations of a SPICE-referenced TPT/ZDF core, retaining earlier-stage nonlinearities preserves harmonic behavior better than retaining later-stage nonlinearities alone. Beyond the specific comparisons reported here, the framework provides a reproducible basis for evaluating fidelity–cost trade-offs when nonlinear structure must be simplified.
Download Evaluating Dynamic Range Compressor Models Using Control-Voltage Measurements: An Approach and Dataset The quantity that defines the behavior of a dynamic range compressor is the time-varying gain applied to the signal as a function of the input level. However, models of these devices are typically evaluated using proxy metrics because isolating the gain reduction signal from the audio input–output data included in existing datasets creates an ill-conditioned inverse problem. It is unclear how accurately these metrics describe the behavior the model is tasked with emulating, particularly as waveform-based metrics can be influenced by secondary effects introduced by analog processing and capture, even when those effects are inaudible. We investigate a method of evaluation in which the gain-reduction signal produced by a model is measured directly against a gain-reduction control voltage signal produced by the hardware. To evaluate the efficacy of this metric as a learning objective, a gray-box model is trained using loss computed directly over the gain control signals alongside two models trained using common proxy losses. The models trained using proxy losses did not achieve parity with models trained directly on the gain control signal when evaluated with respect to the underlying control trajectory, and the waveform-domain metrics assigned similar errors to models that were clearly separated by the direct metric. To facilitate further exploration of this method of evaluation, we present a Solid State Logic bus compressor dataset that includes the gain control voltage signal captured alongside the audio output.
Download Robust Recovery of Deterministic Timecode Signals Under Analog Degradation This paper presents a controlled evaluation framework for recovering deterministic stereo timecode waveforms used in digital vinyl systems. Three lightweight decoding policies are compared: Fixed-Threshold Symbol Decoder (FTSD), Adaptive-Threshold Symbol Decoder (ATSD), and State-Constrained Temporal Decoder (SCTD). The work provides matched comparison and explicit separation of availability, reference-relative correctness, and temporal behavior across noise, dropout, and clipping conditions.
Download Real-Time Neural Audio on Apple Silicon: Benchmarking Inference Frameworks Under Realistic DAW Contention Neural network models are increasingly deployed in audio plugins across a wide range of applications, including amplifier emulation, effects modeling, and synthesis. This paper evaluates widely used inference options including BNNSGraph, RTNeural, LibTorch, ONNX Runtime, and anira on model architectures commonly used in neural audio plugins. The key contribution is moving beyond isolated benchmarks to evaluate performance under realistic DAW contention, constructing mix sessions with configurable plugin loads. Results show that isolated benchmarks can be misleading, and BNNSGraph proves most robust for convolutional models on Apple Silicon.
Download Benchmarking Integrated GPU Acceleration of Real-Time Neural Audio Inference on Snapdragon This paper investigates whether integrated GPUs on Qualcomm Snapdragon SoCs can accelerate streaming inference of neural audio models. Five models spanning three orders of magnitude in parameter count are benchmarked across three inference approaches (best available CPU, QNN CPU, and QNN GPU). Results reveal when GPU acceleration offers meaningful gains, when per-call overhead negates benefits, and how model size and architecture determine GPU suitability.
Download Perceptually Motivated Alignment and Interpolation of Pitch-Aligned Time-Frequency Representations This paper proposes methods for alignment and interpolation of pitch-aligned time-frequency representations, building on the tonal interval vector. Extensions reformulate it as an invertible operator, enabling alignment via permutation search under perceptually weighted distances and interpolation via optimal transport with a circular formulation respecting harmonic structure. The work also develops a geometric scale representation factorizing scale structure into root, density, and color.