STFT-Domain Least-Distortion Region-of-Interest Beamforming

This paper introduces a new method for maximizing array gain beamforming while focusing on a specific area in space. The main goal is to minimize the distortion of a desired signal when there is uncertainty about its actual direction of arrival. We develop a signal model and establish appropriate constraints and measures. We propose two beamformers that minimize distortion, each optimized for either white noise gain or directivity factor. Both beamformers are controlled by a design parameter that balances array gain and average signal distortion. We conduct extensive simulations to validate our approach. We examine the nature of this parameter across different areas of interest and evaluate its impact on the performance of the proposed beamformers using relevant measures. Finally, we conduct speech-signal experiments in various noisy and reverberant environments and consider different array calibration conditions. Our results show that the proposed approach outperforms recently proposed and traditional methods in terms of speech intelligibility, quality, and mean opinion scores, especially when there are significant deviations in the direction of arrival of the desired signal and when the reverberations are mild.