Multichannel Source Activity Detection, Localization, and Tracking
Pasi E. Pertila, Alessio Brutti, Piergiorgio Svaizer, Maurizio Omologo · 2018
An acoustic scene may consist of several contributions, namely a number of sound sources, point or diffuse ones, in particular human speakers, noise sources, reverberation effects due to reflections on obstacles and walls inside an enclosure. Given the possibility of observing this acoustic scene from different spatial positions, an immediate set of problems can be tackled in order to derive the activity and the location of each sound source. Microphone arrays and distributed microphone networks can be fruitfully deployed for these tasks which are not trivial unless under controlled conditions. In this chapter we provide an overview of the basic notions and the most popular algorithms for speech activity detection and acoustic source localization in multi-microphone recordings. We focus on established solutions, giving insight about emerging trends and most recent developments.