Reproduce “Microphone Geometry Aware Regional Target Speaker Extraction” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Shitong Xu, Niki Trigoni, Andrew Markham

The authors introduce a regional speech extractor that uses microphone audio and coordinates to support different array layouts. Cross-microphone attention and pooling create a fixed-size representation, while an azimuth request guides extraction. In simulated tests, it achieves 9.79 dB SI-SDRi on Dynamic-4, versus 4.68 dB for a four-microphone baseline, but struggles to suppress speakers just outside regions on a tetrahedral array.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.