Skip to main content
Search all SpringerOpen articles
EURASIP Journal on Audio, Speech, and Music Processing
Table 7 Results obtained with ULT+WRE approach compared with the
. Digit Error Rate depending on the weight reestimation rules (MLE and FD) with ULT. 2 300 tests performed on BDSON corpus (clean).
Compact Acoustic Models for Embedded Speech Recognition
Very compact model
Back to article page