Coppens A., Hermen J., Maquil V.
Conference on Human Factors in Computing Systems Proceedings, art. no. 612, 2026
Speech is a common modality for interactive systems and collaboration analysis in multi-user studies. Despite the widespread availability of automated transcription tools, attributing speech reliably to individual participants remains challenging in collocated settings. Yet, microphone type and placement are often overlooked, considered as pragmatic constraints rather than being treated as design variables. In this paper, we present a controlled study evaluating how wearable microphone configurations affect the robustness of target speaker extraction using a lightweight speech detection approach. Three configurations were compared: a shirt-mounted clip-on microphone, the same microphone positioned near the wearer's mouth via a rigid necklace, and a commercial headset. We show that shirt-mounted clip-ons often fail to isolate the wearer's speech from the non-wearer's, whereas near-mouth and headset placements substantially improve robustness. These findings highlight microphone choice as a key design decision for speech-based interactive systems and multi-user studies, providing empirically grounded guidance for researchers and practitioners.
