An expandable software-framework for acoustic monitoring applications

Developers of real-world audio monitoring applications often face similar challenges. When newly designed algorithms and demonstrators are ready to prove their capabilities in an offline setting, the way to a robust real-time application is still long and full of obstacles. Apart from implementing algorithms in a suitable manner, amongst the requirements are multichannel signal capturing with reference to local time, storage and retrieval of detected events, system-fault detection and recovery as well as communication with the outside-world. To meet these requirements in a flexible way, we propose a distributed approach, allowing for computational scalability as well as robustness by proportion of the employed hardware. With today's availability of low-cost, high-speed Ethernet links, transmission of uncompressed multichannel audio-data in real-time has ceased to be unreasonable. By setting up on a suitable middleware, advantages of several software technologies can be combined to form a framework capable of conforming to our requirements. We will give an overview of the system's software architecture, as well as its current implementation and performance. The framework has been successfully deployed in the course of our acoustic tunnel monitoring project AKUT.