Physics-Informed Transfer Learning for Data-Driven Sound Source Reconstruction in Near-Field Acoustic Holography

2025-07-15

AI Summary

To address the poor cross-source generalization of data-driven models in near-field acoustic holography (NAH), this paper proposes a physics-informed transfer learning framework. First, a complex-valued convolutional neural network (CV-CNN) is pre-trained on diverse acoustic source data; then, single-sample fine-tuning is performed under physical constraints derived from the Kirchhoff–Helmholtz integral equation. This enables rapid model adaptation to unseen source types without requiring extensive labeled data from the target source. In experiments on a violin soundboard, the fine-tuned model achieves significantly higher reconstruction accuracy than both the pre-trained CV-CNN and the compressed equivalent source method (C-ESM), reducing key modal error by up to 32%. The core innovation lies in deeply embedding acoustic physical priors into the transfer learning pipeline—thereby synergizing data-driven representation capability with acoustic interpretability—and substantially enhancing the cross-domain robustness and practical applicability of NAH models.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We propose a transfer learning framework for sound source reconstruction in Near-field Acoustic Holography (NAH), which adapts a well-trained data-driven model from one type of sound source to another using a physics-informed procedure. The framework comprises two stages: (1) supervised pre-training of a complex-valued convolutional neural network (CV-CNN) on a large dataset, and (2) purely physics-informed fine-tuning on a single data sample based on the Kirchhoff-Helmholtz integral. This method follows the principles of transfer learning by enabling generalization across different datasets through physics-informed adaptation. The effectiveness of the approach is validated by transferring a pre-trained model from a rectangular plate dataset to a violin top plate dataset, where it shows improved reconstruction accuracy compared to the pre-trained model and delivers performance comparable to that of Compressive-Equivalent Source Method (C-ESM). Furthermore, for successful modes, the fine-tuned model outperforms both the pre-trained model and C-ESM in accuracy.

Problem

• Transfer learning for sound source reconstruction in NAH
• Physics-informed adaptation across different sound sources
• Improving accuracy with pre-trained and fine-tuned models

Innovation

• Transfer learning for sound source reconstruction
• Physics-informed fine-tuning with Kirchhoff-Helmholtz integral
• Complex-valued CNN pre-trained on large dataset