Advancing Direction-Preserving Noise Reduction for Higher-Order Ambisonics Signals: Comparative Analysis and Methodological Refinement

Abstract

This study delves into refining noise reduction techniques for higher-order Am- bisonics (HOA) signals, aiming to preserve directional accuracy amid unwanted noise. Addressing a critical gap in spatial filtering methods, it scrutinizes a direction-preserving approach against beamforming-based strategies and the matrix multichannel Wiener filter. The research explores various methodolo- gies for estimating power spectral densities crucial for noise reduction while proposing a novel technique to protect desired signal reverberation. Assess- ments encompass diverse speech sources in anechoic and reverberant settings, evaluating performance across different noise profiles. This investigation sig- nificantly advances noise reduction tailored for HOA signals, emphasizing the imperative retention of directional fidelity amidst noise interference. The com- parative analysis offers comprehensive insights into these methodologies, their efficacy, and limitations across varied acoustic conditions, contributing substan- tially to enhancing noise reduction strategies for spatially encoded audio signals.