The paper presents a study on an online target sound extraction (TSE) process that uses the similarity-and-independence-aware beamformer (SIBF). The SIBF is a linear method that estimates the target sound more accurately compared to a reference, and this process helps reduce latency. However, the study notes that challenges include potential degradation of accuracy and an increase in the accuracy gap between two algorithms due to the conventional post-process. The paper proposes a novel scaling method based on the single-channel Wiener filter (SWF) to minimize this gap. The study concludes that the online SIBF outperforms the conventional linear TSE, including the minimum mean square error beamformer.
Publication date: 29 Dec 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2312.16449
