Cocoon: On-body Microphone Collaboration for Spatial Awareness

We are now surrounded by multiple microphones, available on various devices around us including wearables. This opens unique opportunities for acoustic sensing applications to enhance the spatial resolution of audio signals on the body. However, the state-of-the-art acoustic sensing applications still mostly utilize a single microphone or a microphone array on a single device. In this paper, we present Cocoon, a case study for on-body microphone collaboration for spatial awareness. Cocoon is a novel wearable system that automatically provides users with situational services based on their location. To this end, it combines spatial profiles from multiple on-body microphones on the fly and identifies a user's location by matching the profile against the pre-registered ones. Our experimental results show that Cocoon outperforms the existing single microphone-based methods, 10.0% points and 21.5% points accuracy increase in the controlled and real-world setup, respectively. Cocoon also improves the robustness to slight movements and orientation changes of microphone, reducing the error rate by 17.5% points.