P1-14: A Dual-Array Beam Cross-Scanning Framework for Source Localization

Guo, Xin*, Zhao, XuDong, Huang, Gongping, Chen, Jingdong, Benesty, Jacob, Cvetkovic, Zoran

Subjects (starting with primary): Source localization and tracking

Presented in Poster Session 1

Abstract:

This paper investigates sound source localization using two spatially separated compact microphone arrays. For a single compact array operating under the far-field assumption, only reliable direction-of-arrival (DOA) information can typically be obtained, making accurate two-dimensional (2D) localization challenging. To address this limitation, a dual-array beam cross-scanning framework is proposed. Within this framework, a frequency-bin-level fusion strategy is developed to jointly integrate spatial responses from the two arrays over a candidate position grid. Specifically, frequency-domain spatial observations are first constructed for each array using a Householder transformation (HT)-based method, which enhances the representation of directional signal components. These observations are then jointly fused at the frequency-bin level to directly form a broadband spatial response on the candidate grid, from which the source position is estimated. Experimental results demonstrate that the proposed method achieves consistently strong performance across various conditions.

This PDF is password protected. Enter the proceedings password in the PDF viewer when prompted.