Anonymous - 2001-08-28

Hi,

I 've tried to build an acoustic model with a long audio file but when I use the baum welch algorithm, an error occurs :

error 3768 749 error : backward.c, line 929 : alpha(5.046354e-01) <> sum of alphas*betas (4.995700e-01) in frame 3926

So, I split my long audio file (1.2min long) into severals files (0.03min).
The baum welch algorithm works on these files.

Is it normal ?
Can I use the training on long audio file ?

Thank you for your help....