AI

Ambisonics soundfield navigation using directional decomposition and path distance estimation

Abstract

We propose a method of ambisonic soundfield navigation based on optimized directional decomposition that extracts ordered directional signals and estimates signal path distances using a correlation matrix. The method allows for room-scale soundfield navigation using a single ambisonic microphone signal. In comparison to other source-based navigation approaches, our approach more accurately handles differences between direct sources and reflections, which results in more plausible behavior when translating in spaces with many observable reflection points. Further, we show our method is efficient in computation and memory and integrates easily with existing ambisonic binaural rendering solutions. We evaluate the quality and accuracy of our translation method using synthetic signals convolved using image-source-method generated B-format room impulse responses.