A Context Clustering technique for Average Voice Models (1)

3. Shared decision Tree Context Clustering

3.1 Training of Average Voice Model

1. A block diagram of the training stage of average voice model using the proposing technique is shown in Fig. 2. First, context dependent models without context clustering is separately trained for respective speakers toderive a decision tree for the context clustering common to these speaker dependent models. Then, the decision tree, which we refer to asA shared decision tree, is constructed usingAn algorithm described in Sect. 3.3From the speaker dependent models. Finally,All speaker dependent models is clustered using the shared decision tree. A Gaussian PDF of Average voice model is obtained by combining all speakers ' Gaussian PDFs at every node of the tree.After the reestimation of parameters of the average voice model using training data of all Speak-ers, State duration distributions are obtained for each speaker. Finally, State duration distributions of the Av-erage voice model was obtained by applying the same procedure.
   1. Fig. 2 The techniques proposed to train the average voice model
   2. Context-sensitive model, without context clustering, is trained separately

A Context Clustering technique for Average Voice Models (1)