Acoustic Model with Multiple Lexicon Types for Indonesian Speech Recognition

Table 13

The performance of the acoustic models using 46,550 unvalidated utterances.

Lexicon type	GMM-HMM model (%WER)
	Monophone	Triphone DELTA + DELTA-DELTA	Triphone LDA + MLLT	Triphone SAT
1c-vocab-full	85.77	69.21	60.44	60.45
enmap-1c-vocab-full	86.10	69.65	61.40	61.10
enmap-kv-vocab-full	83.58	73.46	65.22	65.19
kv-vocab-full	82.56	73.38	65.98	65.96