INFO: main.c(162): No lexical transcripts provided
INFO: corpus.c(1240): Will process all remaining utts starting at 0
INFO: main.c(266): Will produce FEAT dump
INFO: main.c(420): Writing frames to one file
stat_retry(d:/Mine/cp/time/feat/mbuh/aerob
.mfc) failed
ERROR: "corpus.c", line 1538: MFCC read failed. Retrying after sleep...
stat_retry(d:/Mine/cp/time/feat/mbuh/aerob
.mfc) failed
ERROR: "corpus.c", line 1538: MFCC read failed. Retrying after sleep...
......
ERROR: "corpus.c", line 1538: MFCC read failed. Retrying after sleep...
FATAL_ERROR: "corpus.c", line 1542: Failed to get the files after 10 retries of getting MFCC(about 30 seconds)
Thu May 4 13:15:25 2006
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
thanks for all your help. I've already try like your's way jerry and eric, but the result is still same. It cann't found the *.mfc file location.
I also try to using an4 and rm1 database. By thoose database, SphinxTrain can recognize the path location of the .mfc files. Then I just try to add my .mfc files, and other configuration to both database, rm1 and an4, and Thank God it works. It can recognize my mfc files. But in my own database, still doesn't works.
I think it doesn't matter as long as my mfc files can be recognized. ^_^
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
There is something that i don't understand, when just verify using verify_all.pl script, there is no problems. But, when i using the feature files using agg_seg.pl thoose error message appear. Is there something wrong?
By the way, this is the content of my <file>.fileids:
...
mbuh/ada
mbuh/aerob
...
and also the content of my <file>.transcription:
...
<s> ADA </s> (ada)
<s> AEROB </s> (aerob)
...
thanks.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2006-05-11
I have never been able to get SphinxTrain to work with directories in the x.fileids file. I believe it's easiest if all the files are in a single directory, and x.fileids looks like:
...
ada
aerob
...
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I also added -outputfullpath to output the full path in the fileid file in the log output of bw. The reason being that the last component of the path may not uniquely identify an utterance in the corpus (making easy association of bw 'utt>' output lines to utterances very difficult in some corpora).
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
please help me solve this problem. I'm using agg_seg.pl script, there is error message that i can't understand, please read this log:
/cygdrive/d/Mine/cp/time/bin/agg_seg \ -segdmpdirs d:/Mine/cp/time/bwaccumdir/time_buff_1 \ -segdmpfn d:/Mine/cp/time/bwaccumdir/time_buff_1/time.dmp \ -segtype all \ -ctlfn d:/Mine/cp/time/etc/time_train.fileids \ -cepdir d:/Mine/cp/time/feat \ -cepext mfc \ -ceplen 13 \ -agc none \ -cmn current \ -feat 1s_c_d_dd \ -stride 1
...
INFO: main.c(162): No lexical transcripts provided
INFO: corpus.c(1240): Will process all remaining utts starting at 0
INFO: main.c(266): Will produce FEAT dump
INFO: main.c(420): Writing frames to one file
stat_retry(d:/Mine/cp/time/feat/mbuh/aerob
.mfc) failed
ERROR: "corpus.c", line 1538: MFCC read failed. Retrying after sleep...
stat_retry(d:/Mine/cp/time/feat/mbuh/aerob
.mfc) failed
ERROR: "corpus.c", line 1538: MFCC read failed. Retrying after sleep...
......
ERROR: "corpus.c", line 1538: MFCC read failed. Retrying after sleep...
FATAL_ERROR: "corpus.c", line 1542: Failed to get the files after 10 retries of getting MFCC(about 30 seconds)
Thu May 4 13:15:25 2006
thanks for all your help. I've already try like your's way jerry and eric, but the result is still same. It cann't found the *.mfc file location.
I also try to using an4 and rm1 database. By thoose database, SphinxTrain can recognize the path location of the .mfc files. Then I just try to add my .mfc files, and other configuration to both database, rm1 and an4, and Thank God it works. It can recognize my mfc files. But in my own database, still doesn't works.
I think it doesn't matter as long as my mfc files can be recognized. ^_^
The path of mfcc is wrong. Try to check it. -Arthur
check the path in etc/<some_name>.fileids file,
make sure d:/Mine/cp/time/feat/mbuh/aerob
.mfc file exist in this path.
i've already check the path and it's true.
There is something that i don't understand, when just verify using verify_all.pl script, there is no problems. But, when i using the feature files using agg_seg.pl thoose error message appear. Is there something wrong?
By the way, this is the content of my <file>.fileids:
...
mbuh/ada
mbuh/aerob
...
and also the content of my <file>.transcription:
...
<s> ADA </s> (ada)
<s> AEROB </s> (aerob)
...
thanks.
I have never been able to get SphinxTrain to work with directories in the x.fileids file. I believe it's easiest if all the files are in a single directory, and x.fileids looks like:
...
ada
aerob
...
I recently modified bw to accept the command line argument -fullsuffixmatch to allow transcript files with lines like:
ZAPHOD BEEBLEBROX (hitchhiker-2343/Speaker003/000-MOO)
to match fileid files with lines like:
I also added -outputfullpath to output the full path in the fileid file in the log output of bw. The reason being that the last component of the path may not uniquely identify an utterance in the corpus (making easy association of bw 'utt>' output lines to utterances very difficult in some corpora).
My battle scars say to check your fileid's file to make sure that it has only '\n' at the end of the line rather than '\r\n'