Hello every one.
I'm trying to use the pocketshynx librairy for a while in C, and i'm still
not able to recognize anything...
I try to make some test from the goforward sample.
So, when i use the goforward raw file from the package, the sentence is
recognized. But, if I try to record my own one, with the same sentence, but
form my microphone, i get the following message:
WARNING: "ngram_search.c", line 1087: </s> not found in last frame, using <s> instead
INFO: ngram_search.c(1137): lattice start node <s>.0 end node <s>.0
INFO: ps_lattice.c(1228): Normalizer P(O) = alpha(<s>:0:242) = -537889843
Is someone can help me, or give me a simple previous code sample which can
record one unique given sentence and recognize it from the microphone (using
alsa)
Thanks.
Ps: i've theoreticals and practicals basics in speech recognition, using HTK,
Julius, and Julian, but i want use Sphinx because of the C api.
Thanks again!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Yes i guess it's something wrong about the recording format.
But I already use 16khz with one channel. After some research i found that i
have to use Big Endian encodage so i tried (SND_PCM_FORMAT_S16_BE) signed
16bits Big endian but it still doesn't work...
When i import in audacity the original record and mine, with the sames
parameters everything seems ok...
Thanks for you help
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
And of course it's always easier to share the audio than to talk about it. If
you still have troubles, just upload audio to a public sharing resource and
give a link here.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Yeaah! I've something now! It was because of the Big Endian!
Now i've to improve my english accent to have better results ( i'm french ) :)
Thanks a lot
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello every one.
I'm trying to use the pocketshynx librairy for a while in C, and i'm still
not able to recognize anything...
I try to make some test from the goforward sample.
So, when i use the goforward raw file from the package, the sentence is
recognized. But, if I try to record my own one, with the same sentence, but
form my microphone, i get the following message:
Is someone can help me, or give me a simple previous code sample which can
record one unique given sentence and recognize it from the microphone (using
alsa)
Thanks.
Ps: i've theoreticals and practicals basics in speech recognition, using HTK,
Julius, and Julian, but i want use Sphinx because of the C api.
Thanks again!
You are recording in the wrong format.
File must be 16khz 16bit mono MSWAV file. That is one of the practical basics
using HTK, Julius or any other ASR engine.
Yes i guess it's something wrong about the recording format.
But I already use 16khz with one channel. After some research i found that i
have to use Big Endian encodage so i tried (SND_PCM_FORMAT_S16_BE) signed
16bits Big endian but it still doesn't work...
When i import in audacity the original record and mine, with the sames
parameters everything seems ok...
Thanks for you help
Audio needs to be little endian. Not sure where you got the idea about the big
endian.
And of course it's always easier to share the audio than to talk about it. If
you still have troubles, just upload audio to a public sharing resource and
give a link here.
Yeaah! I've something now! It was because of the Big Endian!
Now i've to improve my english accent to have better results ( i'm french ) :)
Thanks a lot