Estimation of acoustic signals in reverberant conditions via the characteristic function
Tran Huy Dat, Fumitada Itakura · The Journal of the Acoustical Society of America · 2003
In many circumstances at reverberant conditions, acoustic signals are mixed by the multiple reflections and scattering phenomena. Recently the independent component analysis (ICA) has gained great advantages for processing this class of signals. Many implementations for both temporal and temporal-frequency, stationary and nonstationary, second order and high order statistics are given. However in these cases, the dynamic structure described by the changing of the time of sources number was not considered. In this paper first we proposed a generalized stochastic model allowing describing both a static and dynamic structure of sources. Then the joint characteristic function is analytically obtained. Second, we examine the estimation of an empirical characteristic function for given models.