I found the broblem you can find the log file in cp_mean the mean n_cb (== 216) it means phone * satahmm you must check the phone number and hmm stat ,
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi
i am trying to train sphinx 3 using my own feature extraction routines. i am getting the following error:
FATAL_ERROR: "main.c", line 352: src mean n_cb (== 216) inconsistent w/ ts2cb mapping 126
the training corpus is timit, and i am using force alignment for training.
In previous module, the baum-welch went through 7 iterations, and the final iteration had a lots of " final state not reached" errors:
si2206 493 0 228 44 ERROR: "backward.c", line 400: final state not reached
I think that I may be better off training with phone aligned transcriptions ? That leads to my next question:
how does one write phone-aligned transriptions ?
thank you
rgds
I have slove the problem ,You must check your phonelist again to find duplicate phone
I found the same problem in windows system but no in linux system you can try again in linux system
I found the broblem you can find the log file in cp_mean the mean n_cb (== 216) it means phone * satahmm you must check the phone number and hmm stat ,