Language Recognition Using Latent Dynamic Conditional Random Field Model with Phonological Features

Table 9

Error of language classifiers trained on LDCRF, CRF, and GMM models using PFs and classifiers trained on GMM using PLP (averaged across 28 pairs).

Model	Error (%)
LDCRF (PF)	17.46
CRF (PF)	19.11
GMM (PF)	30.63
GMM (PLP)	52.28

The least error result is displayed in bold.