Figure 3From: Homogenous ensemble phonotactic language recognition based on SVM supervector reconstruction SVM supervector of an utterance (a) before and (b) after perturbation. Back to article page