Acoustic Model with Multiple Lexicon Types for Indonesian Speech Recognition

Table 6

The performance of the acoustic models using the unvalidated dataset.

Lexicon type	GMM-HMM model (%WER)				TDNNF (%WER)
	Monophone	Triphone DELTA + DELTA-DELTA	Triphone LDA + MLLT<	Triphone SAT
vocab_1_char	76.08	56.06	46.09	47.05	29.62
enmap_vocab_1_char	75.93	54.92	46.15	47.04	29.77
norm_1c_vocab_full	76.18	56.08	46.56	46.36	29.89
enmap_norm_1c_vocab_full	76.87	55.57	46.25	46.17	29.79
kv_vocab_full	71.19	58.61	47.82	48.81	29.49
enmap_kv_vocab_full	72.05	58.55	48.29	47.64	29.41