Re: [SoX-users] Record of Voice

[Jan S. <ha...@st...>]

On Jan 30 11:04:17, onk...@gm... wrote:
> Speech-to-text engine -->
> http://mikepultz.com/2011/03/accessing-google-speech-api-chrome-11/
> 
> When i speak and the engine gives a string back, my different modules check
> if special keywords are in the string (like "remind me" or "start pc one").
> Then special action start (send a WOL packet or start a remind-me-skript)
> and i start a voiceoutput with a text-to-speech engine. But this all works
> fine! :)
> I use this micro for that:
> http://www.amazon.com/Audio-Technica-ATR-4697-Omnidirectional-Condenser-Microphone/dp/B002GR7INW/ref=cm_cr_pr_product_top
> 
> With the command "rec -c 1 -r 16000 -t flac temp/rec$i.flac silence 1 0.1
> 7% 2 1.0 5%" the system stops nearly every 10 seconds the record in my
> environment and would send it to the internet.

Why would it stop every ten seconds? It stops recording when
the 'silence' parameteres are met, which can be after 10 seconds
or after 5 minutes, depending on the input.

If I am reading your parameters riht, you are instructing SoX
to start recording upon spotting 0.1 seconds of 7%-non-silence,
and stop recording after two seconds of 5%-silence.

So start talking and talk for a minute, without making a 2s pause.
It will record a minute.

> The robot should run 24/7 so
> it is not an alternative to send every 5 to 10 seconds files without
> content to the internet. Because of that i build the
> "Rough frequency"-Filter.

Whatever you record is the "content".
Tuning it to only record your speech is precisely the matter of
setting up the paramaters of 'silence' and possibly other effects.

> >You can use a band filter to only record the frequency range
> >of the human voice, and then use the compander to make
> >the signal somehow uniformly strong
> 
> Can you give me an example command for this? Maybe this could be an
> alternative to my "Rough frequency"-Filter when just voice will be
> recorded?!

Read about the 'bandpass' and 'compand' effects in SoX's manpage.
You will also find examples there.

See http://en.wikipedia.org/wiki/Human_voice
and try SoX's spectrogram effect to see what
the frequency band of your voice is
(typically within 0Hz - 4kHz)