Figure 1From: Compact Acoustic Models for Embedded Speech RecognitionAn overview of the proposed architecture. For state , the state-dependent GMM (GMM) is obtained by applying the transformation function to the state-independent GMM.Back to article page