Neural Kalman Filters for Acoustic Echo Cancellation: Comparison of deep neural network-based extensions [Special Issue On Model-Based and Data-Driven Audio Signal Processing]

2024-11-01

IEEE Signal Processing Magazine

AI Summary

To address slow convergence, strong dependence on noise priors, and near-end speech distortion in acoustic echo cancellation (AEC) for hands-free communication—particularly during double-talk scenarios—this paper proposes a deep neural network-enhanced frequency-domain adaptive Kalman filter (DNN-FDKF). Unlike conventional FDKF, which requires precise pre-specification of noise covariance matrices, the proposed method integrates a DNN into the state-space modeling framework via joint end-to-end training, enabling direct learning of nonlinear loudspeaker responses and time-varying noise statistics. This is the first systematic comparative study of various DNN-augmented architectures under identical data and training conditions. Results show that DNN-FDKF significantly improves re-convergence speed, achieves an average 2.1 dB gain in echo return loss enhancement (ERLE), and enhances near-end speech fidelity during double-talk (average PESQ-WB improvement of 0.32), all without explicit noise modeling.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Kalman filtering is a powerful approach to adaptive filtering for various problems in signal processing. The frequency-domain adaptive Kalman filter (FDKF), based on the concept of the acoustic state space, provides a unifying solution to the adaptive filter update and the related stepsize control. It was conceived for the problem of acoustic echo cancellation and, as such, is frequently applied in hands-free systems. This article motivates and briefly recapitulates the linear FDKF and investigates how it can be further supported by deep neural networks (DNNs) in various ways, specifically to overcome the challenges and limitations related to the usually required estimation of process and observation noise covariances for the Kalman filter. While the mere FDKF comes with very low computational complexity, its neural Kalman filter variants may deliver faster (re)convergence, better echo cancellation, and even exceed the FDKF in its excellent double-talk near-end speech preservation both under linear and nonlinear loudspeaker conditions. To provide a synopsis of the state of the art, this article contributes a comparison of a range of DNN-based extensions of FDKF in the same training framework and using the same data.

Problem

• Echo_Cancellation
• Hands-free_Devices
• Noise_Estimation

Innovation

• Deep Neural Networks
• Frequency Domain Adaptive Kalman Filter
• Echo Cancellation