Hey everyone 

I just develop a recognizer and it perfomrs well using the live audio but when I test its performance using test data off-line I got degraded performance. what do you think the problem might be?

here is the configuration file I used for trainig 

SOURCEKIND = WAVEFORM

SOURCEFORMAT = WAV

SOURCERATE = 625

TARGETFORMAT = HTK

TARGETKIND = MFCC_0_D_A

WINDOWSIZE = 250000.0 

TARGETRATE = 100000.0 

SAVECOMPRESSED= T

SAVEWITHCRC = T

NUMCEPS = 12

USEHAMMING = T

PREEMCOEF = 0.97

NUMCHANS = 26 

CEPLIFTER = 22 

ENORMALISE = F

--- (Edited on 6/25/2014 8:08 am [GMT-0500] by Visitor) ---