Database of Simulated Room Impulse Responses for Acoustic Sensor Networks Deployed in Complex Multi-Source Acoustic Environments

In this work we present a large set of simulated room impulse responses for a multi-room apartment. The simulated apartment models a real vacation apartment for which a recorded set of audio data has already been made available in the context of the DCASE challenges. The impulse responses were rendered using a dense grid of sources and receivers by means of a hybrid auralization algorithm based on a low-order image-source method and deterministic cone tracing. The proposed data set can be used to generate a wide variety of acoustic scenes which, in turn, can benefit numerous data-demanding machine-learning algorithms. An example application is provided in the form of an unsupervised clustering algorithm that groups microphone nodes around dominant sound sources across the entire apartment.