Skip to main content

Table 3 DER and acoustic model size according to the number of Gaussian components per state (context-free models). The acoustic vectors are composed of 13 coefficients (12 PLP plus energy). isolated-digit recognition tests were performed on the corpus BDSON (digit/clean).

From: Compact Acoustic Models for Embedded Speech Recognition

No. of gauss/state No. of parameters DER
2 5 832 4,96%
4 11 664 4,43%
128 373 248 4,52%
full 1 092 096 0,96%