TL;DR
This thesis investigates the application of deep learning techniques to overcome the limitations of conventional ANC systems and proposes a novel Stacked Autoencoder (SAE)- based ANC framework, trained to estimate the optimal anti-noise signal required for noise cancellation.
A feedback-guided DNN-based controller fusion framework for robust fixed-parameter ANC that combines a causal WaveNet controller with a feedback-guided mixture-of-experts (MoE) module, where a gating network estimates the weights of multiple pre-trained FIR experts according to the current acoustic condition.
A temperature-perceptive SFANC (TP-SFANC) approach is proposed that employs a lightweight one-dimensional convolutional neural network (1D CNN) trained using a multi-task learning strategy to dynamically select the optimal control filter.
This conceptual study provides a thorough explanation of ANC technology spanning more than 90 years of development, from Paul Lueg's early 1936 patent to the most recent deep learning methods, and shows that deep learning approaches are the ANC technology of the future.
Deep learning-based active noise control (ANC) algorithms demonstrate superior potential over traditional methods in addressing nonlinear distortion. Although recent deep learning approaches incorporating Volterra Neural Networks (VNNs) have been optimized to tackle nonlinearities, there remains room for further improvement: (1) utilizing element-wise addition in skip connections across different feature processing stages increases the risk of feature aliasing or suppression; (2) directly introducing higher-order terms of the Volterra series is prone to causing overfitting; and (3) models trained under a singular nonlinear condition struggle to adapt to real-world scenarios with varying nonlinearities. To address these issues, this letter proposes a novel time-domain ANC framework. While retaining the modeling capabilities of WaveNet and VNNs, the proposed framework integrates the U-shaped structure and incorporates an adaptive gating mechanism for the higher-order Volterra terms. The proposed method is compared with a state-of-the-art deep learning framework. Additionally, the model is evaluated under several controlled nonlinear conditions, and comprehensive ablation studies are conducted. Simulation results demonstrate that the proposed algorithm outperforms existing end-to-end representative Deep Neural Network (DNN) methods, and ablation studies confirm the effectiveness of the proposed modules.
Distributed multichannel active noise control (DMCANC) has emerged as a scalable framework for large-area noise reduction, where multiple nodes operate local single-channel ANC controllers and exchange essential information to achieve global control. A key limitation of existing DMCANC implementations lies in their reliance on zero or random initialization, which leads to slow convergence of adaptive filters and restricts the efficiency of internode collaboration. To address this issue, this paper introduces a model-agnostic meta-learning (MAML) based initialization strategy for DMCANC. By aggregating heterogeneous acoustic characteristics across nodes-ncluding primary and secondary paths-a MAML framework is trained to learn an initialization that generalizes effectively across distributed ANC systems. The MAML initialization is then deployed to all nodes to improve convergence speed under both stationary and time-varying noise conditions. Numerical simulations applied on broadband and real-world noise demonstrate that the proposed algorithms achieves substantially faster convergence and improved noise reduction performance compared with conventional DMCANC, highlighting the potential of MAML initialization as an effective method for large-scale ANC.
In recent years, many efforts have been made to supersede classical acoustic echo control (AEC) algorithms with more powerful machine-learned approaches. While surpassing the performance of well-established adaptive filters is very much possible, a remaining challenge is computational complexity. Popular architectures, such as convolutional recurrent networks (CRNs), are by multiple orders of magnitude computationally more expensive than classical signal processing solutions. Scaling down such models is usually straight-forward, but it comes at the cost of a notably reduced performance. We show - to the author's knowledge for the first time in AEC - how these performance drops can be successfully alleviated to a large degree by employing an effective knowledge distillation (KD) process, enabling more potent efficient AEC. Our proposed CGGN16 student AEC models show significantly less near-end speech distortion at only 2% of its teacher's computational complexity, surpass the overall performance of a six times more complex model trained on ground-truth labels, and outperform other AEC-focused architectures from recent literature.