Multichannel mobile audio recordings for spatial enhancements and ambisonics rendering

Ambisonics recording allows the playback of a captured 3D audio scene. In many cases, higher-order ambisonics or a set of first-order ambisonics arrays may be used for better results. Virtual ambisonics synthesis and reproduction allows the design and rendering of an audio scene by providing the sound source signals and positions. In this paper, an extended array is proposed, comprising a first-order ambisonics soundfield array, and an arbitrary non-calibrated spatial smartphone-microphone array. The main objective is to investigate whether this extended setup can enhance the spatial resolution of the first-order soundfield array without dramatically increasing the system complexity. The smartphone array provides additional information concerning the relative direction of arrival of different sources, through the calculation of a vector containing the time difference of arrival for all available microphones. A T-SNE visualization is used to show the separability of these vectors that expresses the system's spatial and temporal resolution. Resolution proves to be higher than that of a first-order soundfield array compared to an experiment conducted in previous studies. This supports the idea that the proposed extended array can enhance the resolution of Ambisonic Energy-Based Localization methods.