Figure 4From: Compact Acoustic Models for Embedded Speech Recognition State-dependent transformation by applying ULT followed by WRE. Back to article page