detect silence in raw file?

[wesamoth]

Hi,

I am currently using pocketsphinx to do speech recognition with a robot.
I receive a data flow containing 1 chanel PCM sound (16kHz), and I am able to
record this sound and decode it in batch mode with ps_decode_raw.
But, as it is a streamed sound, I would like to decode it in real time, and
therefore detect silences...

Is it possible to do such a thing? if so, how?

Thanks in advance for your answers.

[Nickolay V. Shmyrev]

For example of continuous decoding check
pocketsphinx/src/programs/continuous.c. Corresponding tool
pocketsphinx_continuous decodes stream from the microphone or from a file
configured with -infile option.

[wesamoth]

You will tell me if I am wrong, but the "infile" option is only available
since 0.7? I should use VS2008 for compatibility reasons, so I can only use
0.6...

However, I will take a look at these functions, just to see if I can adapt it.

[wesamoth]

Thanks!
I just integrated and adapted "ad_file_read" and "recognize_from_file"
functions from the 0.7 version and it works great now!
bonus question: Is there a way to avoid writting into a file and decode
silence directly from a buffer?

[Nickolay V. Shmyrev]

Is there a way to avoid writting into a file and decode silence directly
from a buffer?

You can use sphinxbase AD API to decode from buffers too. It's not different
from reading from a file.

[wesamoth]

Perfect! I had some problems adapting all that stuff but in the end, it works
perfectly.

Thanks for your answers!