Fig. 4From: Scaled norm-based Euclidean projection for sparse speaker adaptationIllustration of a EPL1 and b SNEP-based sparse speaker adaptation from the same ML mean vectorBack to article page