We are trying to build an application with PocketSphinx on Linux (Specifics below ).
Before writing, we searched the web and all forums to find a solution but without success.
Here are our issues:
We’ve followed all the tutorial step by step to correctly install and run PocketSphinx.
However, when we run the program “pocketsphinx_continuous” using the computer microphone, after the listening phase, the results of the listening are always words like:
-“hmmm”
-“mmh”
“hoo” or other similar non-sense phonemes
We used both all the resources you provided on your github (your language model, your acoustic model and CMUDict for US English ... NOT the turtle language : D ) and also created by us but the result is the same. We've also tried to record the input file with the right specifics (Which should be 16KHz, 16 bit PCM, monochannel and pocketsphinx_continuous asked for .wav file...so we feed it with a wav file ) but sadly "MMH" kept ruling the scene.
We tested the turtle language version too, which only works with the provided audio file (Yes, we have said "Go forward ten meters" eheheh ).
We know for sure that the microphone works because we tested it in a lot of various ways.
We have expirienced a problem with ALSA/OSS which required to install pulse audio and to add "aoss" at the head of the pocketsphinx_continuous command ( So this one was solved, but listed just to tell all the story ).
-Have you ever encountered the same problem?
-Is there something we can do to fix the problem?
-Are we doing something wrong?
We are eager to use your valuable program and we look forward to hear from you!
Thanks for your attention,
Kind of regards
The Elly's TEAM
System Info:
We use Linux Debian 8.7 Jessie ( With kernel linux 4.9.0-040900-generic )
Mate 1.8.1 As DE
Memory 2.2 GiB
6 Cores Intel core i7-3610QM @ 2.30 GHz (Base clock, up to 3.6 GHz )
Sadly we have the requirement to stay inside a VE, we use VMWare to virtualize the OS.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
We have expirienced a problem with ALSA/OSS which required to install pulse audio and to add "aoss" at the head of the pocketsphinx_continuous command ( So this one was solved, but listed just to tell all the story ).
You need to install pulseaudio-dev package and then recompile and reinstall sphinxbase instead of using aoss handler.
We've also tried to record the input file with the right specifics (Which should be 16KHz, 16 bit PCM, monochannel and pocketsphinx_continuous asked for .wav file...so we feed it with a wav file ) but sadly "MMH" kept ruling the scene.
You can share the problematic wav file, the command line you use and the pocketsphinx output to get help on this issue.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
We still encounter the same problem, both with speaking to the microphones or providing an audio file as input : results of the listening are always phonemes like “hmm” or “mm”.
See photo attached (output is in yellow/orange color)
Are we using the right and updated resources?
Is there something we can do to fix the problem?
We would be grateful if you could help us on this issue.
Thanks in advance for your patience.
Best regards.
Team Elly
Hello!
We are trying to build an application with PocketSphinx on Linux (Specifics below ).
Before writing, we searched the web and all forums to find a solution but without success.
Here are our issues:
We’ve followed all the tutorial step by step to correctly install and run PocketSphinx.
However, when we run the program “pocketsphinx_continuous” using the computer microphone, after the listening phase, the results of the listening are always words like:
-“hmmm”
-“mmh”
We used both all the resources you provided on your github (your language model, your acoustic model and CMUDict for US English ... NOT the turtle language : D ) and also created by us but the result is the same. We've also tried to record the input file with the right specifics (Which should be 16KHz, 16 bit PCM, monochannel and pocketsphinx_continuous asked for .wav file...so we feed it with a wav file ) but sadly "MMH" kept ruling the scene.
We tested the turtle language version too, which only works with the provided audio file (Yes, we have said "Go forward ten meters" eheheh ).
We know for sure that the microphone works because we tested it in a lot of various ways.
We have expirienced a problem with ALSA/OSS which required to install pulse audio and to add "aoss" at the head of the pocketsphinx_continuous command ( So this one was solved, but listed just to tell all the story ).
-Have you ever encountered the same problem?
-Is there something we can do to fix the problem?
-Are we doing something wrong?
We are eager to use your valuable program and we look forward to hear from you!
Thanks for your attention,
Kind of regards
The Elly's TEAM
System Info:
We use Linux Debian 8.7 Jessie ( With kernel linux 4.9.0-040900-generic )
Mate 1.8.1 As DE
Memory 2.2 GiB
6 Cores Intel core i7-3610QM @ 2.30 GHz (Base clock, up to 3.6 GHz )
Sadly we have the requirement to stay inside a VE, we use VMWare to virtualize the OS.
This is a wrong approach
You should have read
https://cmusphinx.github.io/wiki/faq/#q-failed-to-open-audio-devicedevdsp-no-such-file-or-directory
You need to install pulseaudio-dev package and then recompile and reinstall sphinxbase instead of using aoss handler.
You can share the problematic wav file, the command line you use and the pocketsphinx output to get help on this issue.
Hello Nickolay,
Thank you for your reply.
We followed your tips reinstalling pulseaudio-dev package and sphinxbase and now, the test 'pocketsphinx_continuous -inmic yes’ works.
However, running “pocketsphinx_continuous” using:
the language model present in the PocketSphinx 5prealpha folders;
the dictionary retrieved here:
https://github.com/cmusphinx/cmudict
and the acoustic model retrieved here:
https://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English/en-70k-0.1.lm.gz/download
We still encounter the same problem, both with speaking to the microphones or providing an audio file as input : results of the listening are always phonemes like “hmm” or “mm”.
See photo attached (output is in yellow/orange color)
Are we using the right and updated resources?
Is there something we can do to fix the problem?
We would be grateful if you could help us on this issue.
Thanks in advance for your patience.
Best regards.
Team Elly
This dictionary uses wrong phoneset probably if you included it as is. Try first with the models included in pocketsphinx (model folder).