Efficient parameter transcoding scheme for interactive spatial audio communication

Directional Audio Coding (DirAC) is a well-proven technique for recording spatial sound and efficiently coding it into one or very few audio channels accompanied by parametric side information. Therefore, it is suited for teleconferencing featuring spatial rendering of distributed sources. In teleconferences with more than two attending parties, additional rendering capabilities are desired to (a) spatially distribute each party for better intelligibility of single speakers especially in situations of multiple active sources, (b) adjust levels of individual sources to the requirements of given listening preferences, and to (c) align acoustic with visual cues. MPEG Spatial Audio Object Coding (SAOC) provides this required functionality. Originally, SAOC was designed for having single separated audio objects, i.e., their signals as inputs. In this contribution we propose {DirAC} in acoustic front-end processing for SAOC, where directional filtering in DirAC's parameter domain is used to separate single sources from an acoustic mixture. The paper presents a close look at an efficient transcoding of the parameters of the two considered techniques.