Blind source extraction for robust speech recognition in multisource noisy environments

Efficient online target speech extraction using DOA-constrained independent component analysis of stereo data for robust speech recognition

“…Recently, Nesta and Matassoni [11,12] presented a constrained ICA method based on a semi-blind source separation framework [25]. In the semiblind source extraction (SBSE), a separating matrix constraint is imposed to force the first output to give a target source signal, while the others provide the remaining noise source signals [11,12]. Unfortunately, for estimated (inaccurate) target mixing parameters, no adaptation of the first column of an adaptive separating matrix can provide deteriorated target speech, whereas moderate adaptation does not necessarily guarantee a target source signal in the first system output [12].…”

“…[14,15]). Recently, Nesta and Matassoni [11,12] presented a constrained ICA method based on a semi-blind source separation framework [25]. In the semiblind source extraction (SBSE), a separating matrix constraint is imposed to force the first output to give a target source signal, while the others provide the remaining noise source signals [11,12].…”

“…Some of these techniques can be used in preprocessing for robust speech recognition (e.g. [10][11][12] method that expresses multivariate data as linear combinations of statistically independent random variables, has attracted considerable interest because of its successful performance in many BSS applications (e.g. [8,9,13]).…”

Audio/video supervised independent vector analysis through multimodal pilot dependent components

“…Performance were evaluated by computing both the Noise-to-Speech ratio improvement (NSRi) at the noise output, and the Signal-to-Distortion ratio improvement (SDRi) at the speech output. Indeed, it should be noted that the scenario is highly underdetermined and a complete good speech extraction system should make use of both speech and noise estimates [18] [19]. Fig.…”

Robust audiovisual speech recognition using noise-adaptive linear discriminant analysis

“…The STFTXin(k, t) was subjected to a speech enhancement algorithm with Wiener gain, the decision directed approach for the estimation of the a-priori SNR ξ(k, t) [1], and an improved minima controlled recursive averaging (IMCRA) for estimating the noise power N (k, t) [27]. The estimated noise power N (k, t) was weighted with gain-factor ξ(k, t)/(1 + ξ(k, t)) to arrive at an estimate of the spectral uncertaintiesΣN (k, t) (see Nesta et al [28]). …”