I'm a bit lost in Sphinx. I want to recognize some words recorded by myself (because the demos, using other voices, were not accurate) and get a corresponding text. I read the turorial and I tryed to do the following to achieve my goal:
TRAINING
-compiled the SphinxTrain, using gcc embedded in CIGWIN.
-record some words using windows sound recorder, use SOX to convert them to .sph and put them into <my_db>/wav.
-make .dic, .fille, .fileids, .phone, .transcription files.
-Run scripts_pl/ scripts. (all phases passed)
GENERATING DATA BASE
-copy model_architecture/g_teste_adb.1000.mdef to <my_db>/etc
-copy .dic and .fille to <my_db>/dict
-copy model_parameters/g_teste_adb.cd_cont_1000_8 folder to <my_db>/g_teste_adb.cd_cont_1000_8
-make model.props
-build sphinx4 usin ANT
DECODING
-copy sphinx4/lib/.jar to my project (including my g_teste_8gau_13dCep_16k_40mel_130Hz_30000Hz.jar)
-make g_teste_grammar.gram
-make g_teste_config.xml
-used WavFile.java demo as my main class, replacing the configURL with GTeste.class.getResource("/g_teste_config.xml");
-replacing the wave file with the ones used for training.
-run de WavFile.java
==================================================================================
The output of my Class is always:
Hi.I got the same problem as you,when I change the dictionary to cmudict.0.6d,it is much better.but another problem comes up,it is very very slow.
so I think my be we should build dictionary for a certain grammar,when I just put the words which appeared in the grammar,it shows all SIL again,I am still looking for the solution now.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'm a bit lost in Sphinx. I want to recognize some words recorded by myself (because the demos, using other voices, were not accurate) and get a corresponding text. I read the turorial and I tryed to do the following to achieve my goal:
TRAINING
-compiled the SphinxTrain, using gcc embedded in CIGWIN.
-record some words using windows sound recorder, use SOX to convert them to .sph and put them into <my_db>/wav.
-make .dic, .fille, .fileids, .phone, .transcription files.
-Run scripts_pl/ scripts. (all phases passed)
GENERATING DATA BASE
-copy model_architecture/g_teste_adb.1000.mdef to <my_db>/etc
-copy .dic and .fille to <my_db>/dict
-copy model_parameters/g_teste_adb.cd_cont_1000_8 folder to <my_db>/g_teste_adb.cd_cont_1000_8
-make model.props
-build sphinx4 usin ANT
DECODING
-copy sphinx4/lib/.jar to my project (including my g_teste_8gau_13dCep_16k_40mel_130Hz_30000Hz.jar)
-make g_teste_grammar.gram
-make g_teste_config.xml
-used WavFile.java demo as my main class, replacing the configURL with GTeste.class.getResource("/g_teste_config.xml");
-replacing the wave file with the ones used for training.
-run de WavFile.java
==================================================================================
The output of my Class is always:
Decoding /C:/Documents/TesteSphinx2/src/teste.wav
WAVE (.wav) file, byte length: 55172, data format: PCM_SIGNED 22050.0 Hz, 16 bit, mono, 2 bytes/frame, little-endian, frame length: 27564
RESULT:
the result is always "". Maybe you know something that can cause this.
Thank you. Gustavo Peixoto
one more thing,do not use it in cygwin,I got the blue screen here.
Hi.I got the same problem as you,when I change the dictionary to cmudict.0.6d,it is much better.but another problem comes up,it is very very slow.
so I think my be we should build dictionary for a certain grammar,when I just put the words which appeared in the grammar,it shows all SIL again,I am still looking for the solution now.