yh1008 / speech-to-text

mixlingual speech recognition system; hybrid (GMM+NNet) model; Kaldi + Keras
http://llcao.net/cu-deeplearning17/project.html
70 stars 19 forks source link

fail to create mfcc feature #9

Closed yh1008 closed 7 years ago

yh1008 commented 7 years ago

$ steps/make_mfcc.sh --nj 8 data/train exp/make_mfcc/train mfcc

utils/validate_data_dir.sh: Successfully validated data-directory data/train steps/make_mfcc.sh [info]: segments file exists: using that. run.pl: 8 / 8 failed, log is in exp/make_mfcc/train/make_mfcc_train.*.log

$ less exp/make_mfcc/train/make_mfcc_train.*.log
shows

extract-segments scp,p:data/train/wav.scp exp/make_mfcc/segments.1 ark:- | compute-mfcc-feats --verbose=2 --config=conf/mf
cc.conf ark:- ark:- | copy-feats --compress=true ark:- ark,scp:/home/yh2901/kaldi/egs/codeswitch/mfcc/raw_mfcc_train.1.ark,/
home/yh2901/kaldi/egs/codeswitch/mfcc/raw_mfcc_train.1.scp 
Started at Mon Mar 27 18:20:09 UTC 2017

compute-mfcc-feats --verbose=2 --config=conf/mfcc.conf ark:- ark:- 
copy-feats --compress=true ark:- ark,scp:/home/yh2901/kaldi/egs/codeswitch/mfcc/raw_mfcc_train.1.ark,/home/yh2901/kaldi/egs/
codeswitch/mfcc/raw_mfcc_train.1.scp 
extract-segments scp,p:data/train/wav.scp exp/make_mfcc/segments.1 ark:- 

NI02FAX_0101.flac: ERROR initializing decoder
                   init status = FLAC__STREAM_DECODER_INIT_STATUS_ERROR_OPENING_FILE

An error occurred opening the input file; it is likely that it does not exist
or is not readable.
ERROR (extract-segments[5.0.61~1-37b53]:Read4ByteTag():wave-reader.cc:75) WaveData: expected 4-byte chunk-name, got read errror

[ Stack-Trace: ]
extract-segments() [0x5193f4]
kaldi::MessageLogger::HandleMessage(kaldi::LogMessageEnvelope const&, char const*)
kaldi::MessageLogger::~MessageLogger()
kaldi::WaveData::Read4ByteTag(std::istream&, char*)
kaldi::WaveData::Read(std::istream&, kaldi::WaveData::ReadDataType)
kaldi::WaveHolder::Read(std::istream&)
kaldi::RandomAccessTableReaderScriptImpl<kaldi::WaveHolder>::HasKeyInternal(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, bool)
kaldi::RandomAccessTableReaderScriptImpl<kaldi::WaveHolder>::HasKey(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)
kaldi::RandomAccessTableReader<kaldi::WaveHolder>::HasKey(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)
main
__libc_start_main
_start

WARNING (extract-segments[5.0.61~1-37b53]:Read():feat/wave-reader.h:165) Exception caught in WaveHolder object (reading). 
WARNING (extract-segments[5.0.61~1-37b53]:HasKeyInternal():util/kaldi-table-inl.h:1792) Error reading object from stream 'flac -c -d -s /home/yh2901/kaldi/egs/codeswitch/interview_audio/train/NI02FAX/NI02FAX_0101.flac |'
WARNING (extract-segments[5.0.61~1-37b53]:main():extract-segments.cc:126) Could not find recording NI02FAX_0101, skipping segment NI02FAX_0101_0055711_0060021
WARNING (extract-segments[5.0.61~1-37b53]:Close():kaldi-io.cc:501) Pipe flac -c -d -s /home/yh2901/kaldi/egs/codeswitch/interview_audio/train/NI02FAX/NI02FAX_0101.flac | had nonzero return status 256
yh1008 commented 7 years ago

cause I freaking change the speaker id from 02FAX to NI02FAX, and I forgot to update that in my audio_prep.py script.... shame on me.... instead of /home/yh2901/kaldi/egs/codeswitch/interview_audio/train/NI02FAX/NI02FAX_0101.flac the current (buggy) path is /home/yh2901/kaldi/egs/codeswitch/interview_audio/train/02FA/NI02FAX_0101.flac

I need to have it fixed