Tobias Gburrek, Jörg Schmalenströer, Janek Ebbers, and Reinhold Häb-Umbach
We are going to present the Acoustic Sensor Network SIgnal Generator (ASNSIG) framework¹. ASNSIG enables the simulation of multi-channel audio recordings of various acoustic scenes taking place in a multi-room smart home environment, e.g., cooking and eating. Therefore, the simulated audio signals include not only speech but also various acoustic events occurring in daily life. Beyond the simulation of the source signals ASNSIG enables the simulation of several aspects being typical for an ad-hoc acoustic sensor network (ASN), for example, asynchronously recorded audio signals and the distribution of multiple recording devices in multiple rooms. In addition to that, ASNSIG provides an evaluation data set for many ASN applications, e.g., acoustic source localization, sound event detection and sampling rate offset estimation. Furthermore, we show examples for applications running on the signals generated by ASNSIG, e.g., signal synchronization.
¹ The framework will be released as an open source python package soon.