Spherical array processing for acoustic analysis using room impulse responses and time-domain smoothing

Abstract

Room impulse responses (RIRs) have been widely used for room acoustics analysis. One of the major objectives in room acoustics analysis is characterizing the directions of arrival (DOA) of the direct sound and early room reflections. In the past, spherical microphone arrays have been used to sample the sound field, taking advantage of their spatial symmetry, and various optimal array processing methods have been suggested for DOA estimation. However, these methods fail when the sound field is composed of highly correlated components, such as room reflections, since the cross-spectrum matrix may be of deficient rank. Recently, a preprocessing method incorporating smoothing of the cross-correlation matrix in the frequency domain has been suggested to overcome the rank deficiency. However, when using low-order spherical arrays, this method also fails due to the need to estimate the DOA of a group of reflections simultaneously. In this paper, an alternative time-domain smoothing approach is suggested. The method utilizes the time behavior of the RIR to separate early reflections, together with representation of the signal in the spherical-harmonics domain, to achieve improved performance over current methods. Simulation and experimental studies illustrate the limitations of the former method and the advantages of the suggested method.