The investigation of eXMA method with non-spherical scatters

Abstract

The XMA was a recently presented higher-order ambisonic microphone array which is based on the spherical microphone array (SMA) and equatorial microphone array (EMA) but without a traditional spherical scattering body. Since it is compatible with the EMA, the XMAs can also be designed with the microphones placed on a circumferential contour around the scattering body, which is called the equatorial XMA (eXMA). Compared with the classical SMAs, the eXMA method reduced the required number of microphones significantly since it did not need the microphones to be distributed over the whole surface of the scatterer. The eXMA shows a good application prospect in spatial sound field recording especially when combined with the VR camera to produce a complete panoramic audio-visual experience from a first-person view. However, the eXMA has so far only been evaluated as a headmounted array, i.e. with a human head as the baffle. The performance of eXMA with other shapes of scatterers are unknown. In this work, we used the mesh2hrtf implementation of the boundary element method (BEM) to simulate eXMA calibration measurements for a variety of candidate scatterers including cylinders, cubics and some shapes that are inspired from real VR 360 cameras. We also deformed those shapes and moved up the microphone array to see the influence. Based on those simulations, we identify what spherical harmonic orders can be obtained with what accuracy for a set of convex scattering body geometries that are of relevance in the given context. We demonstrate that the shape of the body is not very critical. The eXMA shows very robust performances with the different shapes of scatterers, some of them even have corners. Reducing the height of the scatterers or moving up the microphone array to the edge will increase the error but the accuracy is still acceptable. The main limitation is the size of the scatters that small bodies do not allow for extracting higher orders at low frequencies. Limitations of the simulation are discussed and at the end we also generate some spatial audio recordings based on the cuboid and the squashed cylinder scatterers.