Acoustic Model with Multiple Lexicon Types for Indonesian Speech Recognition

Figure 7

The %WER of the GMM-HMM models trained using 10,000 unvalidated utterances.