Figure 6From: On pre-image iterations for speech enhancementResults of kernel PCA with normalized pre-imaging (kPCA), PI with SNR-dependent setting of the kernel variance (PIcSNR), the generalized subspace method (Subspace), spectral subtraction (SpecSub), and the MMSE log-spectral amplitude estimator (LogMMSE) in terms of overall perceptual score (OPS), target perceptual score (TPS), interference perceptual score (IPS), and artifact perceptual score (APS) on the test set of the airbone database corrupted by AWGN.Back to article page