I configured sphinxbase using the --enable-fixed option as described in the instructions since I'm running off a Raspberry Pi 3. However, while trying to adapt the acoustic model, the call ./bw -hmmdir... fails with a lot of errors related to a forward.c file. Once I re-installed sphinxbase without the --enable-fixed option, the adaptiation works without errors.Unfortunately, the call to pocketsphinx_continous to actually convert speech to text now fails.
In other words, to adapt my acoutstic model (./bw...), I have to have sphinxbase installed without the --enable-fixed option, but to actually use pocketsphinx I need to re-install sphinxbase with the fixed option enabled.
Anyone know of a way around this?
/sebbesen
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I configured sphinxbase using the --enable-fixed option as described in the instructions since I'm running off a Raspberry Pi 3.
There is no much sense for fixed in Pi3, it has good FPU
In other words, to adapt my acoutstic model (./bw...), I have to have sphinxbase installed without the --enable-fixed option, but to actually use pocketsphinx I need to re-install sphinxbase with the fixed option enabled.
This is correct, you should have floating point installation for adaptation.
Anyone know of a way around this?
Just do not use fixed
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
As I mention in my original post, pocketsphinx_continuous does not seem to work if I did not configure sphinxbase with the --enable-fixed option. In fact, what happes is that only the first word is translated to text. If I re-install sphinxbase with the option enabled the whole sentence is translated.
Any clues?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I configured sphinxbase using the --enable-fixed option as described in the instructions since I'm running off a Raspberry Pi 3. However, while trying to adapt the acoustic model, the call ./bw -hmmdir... fails with a lot of errors related to a forward.c file. Once I re-installed sphinxbase without the --enable-fixed option, the adaptiation works without errors.Unfortunately, the call to pocketsphinx_continous to actually convert speech to text now fails.
In other words, to adapt my acoutstic model (./bw...), I have to have sphinxbase installed without the --enable-fixed option, but to actually use pocketsphinx I need to re-install sphinxbase with the fixed option enabled.
Anyone know of a way around this?
/sebbesen
There is no much sense for fixed in Pi3, it has good FPU
This is correct, you should have floating point installation for adaptation.
Just do not use fixed
Hi Nickolay,
Thank you for your answer.
As I mention in my original post, pocketsphinx_continuous does not seem to work if I did not configure sphinxbase with the --enable-fixed option. In fact, what happes is that only the first word is translated to text. If I re-install sphinxbase with the option enabled the whole sentence is translated.
Any clues?
You need to reinstall sphinxbase first without fixed and then recompile and reinstall pocketsphinx and it will work fine.
That's it. Thanks!