Dynamic Independent Component/Vector Analysis: Time-Variant Linear Mixtures Separable by Time-Invariant Beamformers

Double Nonstationarity: Blind Extraction of Independent Nonstationary Vector/Component from Nonstationary Mixtures -- Algorithms

“…ICE is based on a reduced mixing model parameterization, in which one source is treated as the SOI and the others, which are not subject to separation, as background sources [13]. ICE has been shown equivalent with the minimum entropy approach when the background model is multivariate Gaussian [14] and with ICA when it is multivariate non-Gaussian [15], [16].…”

“…The nonstationary mixing process is less frequently described by a more specific parameterization such as that used in [46], [47]. Recently, semi-time-variant models denoted as CMV and CSV (Constant Mixing/Separating Vector) have been considered in [14], [16], [48]. CMV and CSV are designed for BSE/jBSE in which the SOI is static or moving, respectively, on a dynamic background.…”

“…The nonstationarity is arranged through allowing specific parameters to be changing from interval to interval. In [14], the well-known FastICA algorithm [12] has been extended for CSV and named as FastDIVA (Fast Dynamic IVA).…”

“…This is because the variance of the SOI is allowed to change over the intervals of the CSV mixing model. However, the performance analysis 1 Nonstationary mixing models are sometimes termed as "dynamic models" [14].…”

“…of FastDIVA in [14] as well as the Cramér-Rao bound in [16] has shown that the SOI is not identifiable when its distribution is Gaussian. Dividing data into more (shorter) intervals does not seem very effective since the number of mixing parameters proportionally grows, and, moreover, the Gaussian SOI remains unidentifiable.…”