I have another question. Now my task is very simple, just repeat two phonemes couple of times. I want to implement this task with FSG rather than the phonetic language model. I tried the phonetic FSG, but the generic model does not have single phoneme in the dictionary. How can I implement that?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I use the following command:
pocketsphinx_continuous -infile test/data/goforward.raw -hmm en-us -allphone model/lm/en_US/en-phone.lm.DMP -backtrace yes -beam 1e-20 -pbeam 1e-20 -lw 2.0 -time yes
You see, there is no argument for the dictionary.
Which dictionary should I put these phonemes in? Besides, if I put the phonemes in a dictionary as a word, I need use the "-fsg" option rather than "-allphone", is that right?
Thank you!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Which dictionary should I put these phonemes in? Besides, if I put the phonemes in a dictionary as a word, I need use the "-fsg" option rather than "-allphone", is that right?
Yes
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The method works but the result is not good. It can only recognize the phoneme once (cannot recognize the repetition).
I use the command:
pocketsphinx_continuous -infile MishawakaMen_0eacc4d6087f1b2ca571025676030e0e_1405359270_Baselineatrest_test8_MicrophoneWired.wav -hmm hub4wsj_sc_8kadapt -dict cmu07a.dic -backtrace yes -time yes -bestpath no -fsg utd.fsg
Hi, dear all,
I am trying to do phoneme recognition in pocketsphinx and following the tutorial here: http://cmusphinx.sourceforge.net/wiki/phonemerecognition
But I did not see the phonetic language model "en-phone.lm.DMP". Could anyone tell me where I can download the file, please?
BTW, how can I get the phoneme duration in pocketsphinx?
Thank you very much!
It is part of pocketsphinx sources. If you checkout latest pocketsphinx from github or subversion you'll get it under path specified on the page
With
ps_seg*API, or with-time yesoption to pocketsphinx_continuous.Thank you. They work very well!
I have another question. Now my task is very simple, just repeat two phonemes couple of times. I want to implement this task with FSG rather than the phonetic language model. I tried the phonetic FSG, but the generic model does not have single phoneme in the dictionary. How can I implement that?
You can put corresponding words in the dictionary
Thanks.
I use the following command:
pocketsphinx_continuous -infile test/data/goforward.raw -hmm en-us -allphone model/lm/en_US/en-phone.lm.DMP -backtrace yes -beam 1e-20 -pbeam 1e-20 -lw 2.0 -time yes
You see, there is no argument for the dictionary.
Which dictionary should I put these phonemes in? Besides, if I put the phonemes in a dictionary as a word, I need use the "-fsg" option rather than "-allphone", is that right?
Thank you!
The argument for dictionary is -dict
Yes
Thank you.
The method works but the result is not good. It can only recognize the phoneme once (cannot recognize the repetition).
I use the command:
pocketsphinx_continuous -infile MishawakaMen_0eacc4d6087f1b2ca571025676030e0e_1405359270_Baselineatrest_test8_MicrophoneWired.wav -hmm hub4wsj_sc_8kadapt -dict cmu07a.dic -backtrace yes -time yes -bestpath no -fsg utd.fsg
I put my FSG, model and wav files in the link:
https://drive.google.com/folderview?id=0B8tSafcov_e8cFh0RlpSbXgycTQ&usp=sharing
Could you help me figure out why it fails to recognize the repetition, please?
Thank you.
I still don't find the reason for the result. Will the -remove_noise and -remove_silence options affect the result now?
Thanks.
dictionary should be in alphabetical order, moreover you redefined 1-character words that were already there. find a proper dict for your task here: https://drive.google.com/file/d/0B3I3tp8-IY1fRVBOR0I5dEh1QWs/edit?usp=sharing
Thanks. It works now. :)