Figure 2From: PLDA in the i-supervector space for text-independent speaker verification The EM steps for estimating the loading matrix D for the i-supervector extractor. Back to article page