Sound field diffuseness estimation based on the cross-spectral analysis of ambisonic signals

A new method is presented for estimating the diffuseness of a sound field using the cross-spectral matrix of the Higher-Order Ambisonics signals captured with a microphone array. The diffuseness, which depends on both the isotropy and coherence of the sound waves/sources that constitute the sound field, is estimated by analysing the spread of the eigenvalues of the cross-spectral matrix over a range of frequencies. This method shares some similarities with the COMEDIE technique proposed by Epain and Jin in 2016, the main difference being the use of the cross-spectral matrix instead of the time-domain cross-correlation matrix analysed in COMEDIE. The proposed frequency-domain approach overcomes some of the limitations of the time-domain analysis, such as the overestimation of the diffuseness caused by correlated sources that are not time-aligned at the measurement position.