Citation Context

... deconvolution (BSD) problem. Such a problem emerges, for example, at a cocktail-party being held in an echoic room. Several BSD algorithms were developed in the past. See, for example, the review of =-=[19]-=-. Like ICA, BSD has several applications: (i) remote sensing applications; passive radar/sonar processing [20, 21], (ii) image-deblurring, image restoration [22], (iii) speech enhancement using microp...

Citation Context

... case of instantaneous linear mixtures, e.g., [2]. In the more realistic case of convolutive linear mixtures, time-domain [3], [4] and frequency-domain [5]–[9] methods have been proposed. We refer to =-=[10]-=- for a categorization of existing convolutive BSS methods (see also Section II). Exploiting the nonstationary nature of speech signals, the BSS problem can be solved via the use of second-order-statis...

Citation Context

...s problem, under the narrowband approximation. But, this approach suffers from the ambiguities of arbitrary scaling and permutations of the sources and solving them is a challenging problem in itself =-=[3]-=-. On the other hand, when there is only one source, the problem of blindly estimatingfiltersfromthefilteredversionsofthesourceisawellstudiedproblem. In addition, if the filters are sparsethen the filt...

Citation Context

... Deconvolution (BSD) problem. Such a problem emerges, for example, at a cocktail-party being held in an echoic room. Several BSD algorithms have been developed over the last decades, for a review see =-=[37]-=-. BSD shows potentials in the following areas: (i) remote sensing applications; passive radar/sonar processing [38], [39], (ii) imagedeblurring, image restoration [40], (iii) speech enhancement Depart...

Citation Context

...tuations have two inherent problems: reverberation and source number uncertainty. When we observe a sound in a room, the observation contains reverberation that can bemodeled as a convolutive process =-=[13]-=-. Though methods in the time-frequency domain through a short-time Fourier transform (STFT) are often used to cope with the reverberation, this causes a permutation problem [14]. The permutation probl...

Citation Context

...nerative models: instantaneous and convolutive mixtures (Choi et al., 2005; Ogrady ∗. Martin A. Giese is the corresponding author. c○2011 Lars Omlor and Martin A. Giese.OMLOR AND GIESE et al., 2005; =-=Pedersen et al., 2007-=-). The instantaneous mixture model is defined by the equations: xi(t)= n ∑ j=1 αi js j(t) i=1,···,m. The time-dependent signals xi(t) are approximated by the linear superposition of a number of hidden...

Citation Context

..., [2], and prominent algorithms include [3], [4]. However, the speech mixtures received at an array of microphones are not instantaneous but convolutive; the convolutive blind source separation (BSS) =-=[5]-=- task may be tackled in either the time- or frequency-domain. Notable time-domain solutions are detailed in [6] and [7]. The frequency-domain approach [8] decomposes the signals at the array into narr...

Citation Context

...imate the desired speech signals in the presence of other speech signals or interfering sounds, without the prior knowledge (or with very little information) about the sources and the mixing process (=-=Pedersen et al., 2007-=-). It offers great potentials in many applications such as automatic speech Comon, 1994; Hyvärinen et al., 2009; Comon and Jutten, 2010) has been widely employed and shown to be promising in BSS prob...

Citation Context

...estimation will deteriorate, because the invertibility of H[z] comes to question. We investigated this by generating the polynomial matrix H[z] as follows: [ L∏ ] H[z] = (I − λOiz) H0 (|λ| < 1,λ∈ R), =-=(6)-=- l=0 where matrices H0 and Oi ∈ RD×D were random orthogonal and the λ → 1 limit was studied. ‘Random run’ means random choice of quantities H[z] ands. The AR fit to observation x was performed by the ...