New automatic forward and backward blind sources separation algorithms for noise reduction and speech enhancement

A new regularized forward blind source separation algorithm for automatic speech quality enhancement

“…The proposed RFBSS algorithm corresponds to a FBSS structure controlled automatically (i.e. adaptation of both separation filters w 12 ðnÞ and w 21 ðnÞ) without using neither a MVAD system nor an integrated block that performs an AVAD as given in [1].…”

“…For this purpose, we need to detect the VAD automatically. Two algorithms called VSS-BBSS and VSS-FBSS have been already proposed in [1] that use AVAD. Both of these proposed AVAD algorithms, for the forward and backward structures, are based on structural techniques that allow controlling adequately the step-sizes of the cross-filters.…”

“…The problem of speech enhancement from very noisy observations has became in last year's one of the recent challenges that a big part of scientists and researchers in the world are investigating [1]. This application of speech enhancement is still taking an important place in our daily life because it is used and present in every new teleconferencing systems, such as hands-free telephony, hearing aid, and teleconferencing.…”

“…In this paper, we focus our interest on the BBSS structure and propose a new automatic adaptive algorithm that enhances the speech signal and reduces the acoustic noise without need of any VAD system. The solution given in this paper is algorithmic and not structural as it is given in [1]. The organization of this paper is as follows: in Section 2, we present the principle of BSS.…”

“…However, all of these techniques used a MVAD system that makes them inoperative in practice. The only work that proposed a VAD system with this FBSS and BBSS structure is given in [1]. In this paper, we focus our interest on the BBSS structure and propose a new automatic adaptive algorithm that enhances the speech signal and reduces the acoustic noise without need of any VAD system.…”