Estimating power spectral density for spatial audio signal separation: An effective approach for practical applications

Abstract

The audio signal separation has been extensively studied due to its wide variety of applications. Along with the advancement of processor calculation performance in the last a couple of decades, the research focus has also extended to realising effective and feasible signal separation in practical environments where the problem becomes more challenging. In addition to classical linear filtering techniques, manipulating the spectral amplitude of a signal by applying a postfilter such as the Wiener filter is known to be an effective approach for practical applications. However the postfilter calculation requires the power spectral densities (PSD) of the signals of interest to be estimated beforehand. This article overviews methods for estimating the PSD of signals using spatial characteristics of their sources. Several practical applications that utilise the estimated PSD are introduced with some experimental results demonstrating the potential of the approach for solving challenging problems in practical applications.