Are Identical Sounds Present in Distributed Recordings to Serve as Spatio-Temporal Anchors? A Case Study using the SINS Database

Takao Kawamura, Nobutaka Ono

Abstract: In this study, we explore whether distributed recordings, captured using devices equipped with internally synchronized microphones, contain identical sounds that could provide a reference for spatial and temporal alignment. We define anchor sounds as those originating from the same location with the same waveform, such as a TV power-on sound or a door-closing sound. To examine their presence and potential usability, we test a twostage detection approach that incorporates widely used signal processing techniques, including cross-correlation, to evaluate waveform similarity and time difference of arrival (TDOA) consistency. Our experiments on the SINS database confirm that such identical sounds are present and can be identified using conventional signal processing techniques. While our findings are specific to the SINS database, they suggest similar anchor sounds may also exist in other real-world datasets, potentially enabling applications such as microphone self-localization and synchronization.

External IDs:dblp:conf/apsipa/KawamuraO25a