Platform for Distributed Processing in Dynamic Acoustic Sensor Networks

Haitham Afifi1, Holger Carl1, Tobias Gburrek2, Jörg Schmalenströer2, Janek Ebbers2, and Reinhold Häb-Umbach2
1Hasso-Plattner-Institut, 2Paderborn University

Nowadays there are a lot of smart applications (as seen in smart homes and smart cities) that rely on acoustic signal processing. The most common implementation of such applications is a central (or cloud) implementation, where acoustic data is collected from sensors and processed on a single server. However, this yields many problems such as single-point-of-failure and privacy concerns.

Alternatively, and thanks to the evolution of hardware development of sensors, part of the processing can be off-loaded to multiple local sensors, which can avoid sharing data with third parties (such as cloud service providers). In our previous work, we presented a platform for distributed processing in wireless acoustic sensor networks, while assuming that nodes are static. In this work, we extend the platform to support a dynamic environment; nodes can move or be replaced. Accordingly, our platform handles two main problems: how to hand over a processing between nodes and which nodes should be selected for hand-over.