sox audio recording on a VOX like basis

TheGlenn88 · 02-16-2014, 04:01 PM

Hi,

I'm not new to using Linux, my experience is using the command line for setting up headless servers i.e LAMP, NAS and Media Servers just for personal use. I have little experience with writing scripts but do have an understanding of them.

I have found a very good solution to record audio based on audio level input so it records when sound is inputted and doesn't record silence. The command is:

Code:

rec recording.wav silence 1 5 8% 1 0:00:01 8%

This detects the audio when there is more than 8% sound and records until it hears 1 second of silence.

I need it to record into date and time file formats which i have accomplished in this script. Although it's very very far from ideal and awfully dirty, it works but it's no good for a long term 24/7 solution.

Code:

#!/bin/bash
NAME=`date +%m-%d-%Y_%H-%M-%S`
rec $NAME.wav silence 1 5 8% 1 0:00:02 8%
~/voxrecord.sh

because rec exits when it has finished recording i need a solution to restart it and wait again for more audio, preferebly as a service?

The goal is to have all the recordings in a web interface to be played back as needed.

Can anyone help me get this system to wait-record-save-repeat?

Thanks

TheGlenn88 · 02-16-2014, 04:59 PM

Actually, thinking about the timestamp filename method like in that basic script. it will only take the date and time from when the script was executed. So if it was executed at 10AM and it didn't record anything till 10:45 it will still have a 10AM timestamp so this would not be the way to do it.

Perhaps

Code:

#!/bin/bash

rec temp.wav silence 1 5 8% 1 0:00:02 8%
NAME=`date +%m-%d-%Y_%H-%M-%S`
mv temp.wav $NAME.wav
~/voxrecord.sh

This seems to work more accurately for timestamping

TheGlenn88 · 02-17-2014, 06:45 PM

I have actually progressed further with the script and i think the only thing i need to do is daemonise it and test. the rec command creates the temp.mp3 file even if no sound is recorded, i imagine it's to check that it is able to write before waiting to find out it can't as the recording starts. So the temp file it creates is usually 384 bytes or 768 bytes, perhaps writing headers? i don't know so i've set the script to remove the temp.mp3 file anyway at the start as a failsafe, then wait for audio via rec command. Once it records it saves the file, if it's above 800 bytes it will move to a date-stamped file and add the info to the database, then remove temp file and restart the script. If the script is aborted before anything is recorded it will simply remove temp and exit.

All of the code is borrowed from various sources like stack exchange and modified to work together, no credit goes to me, really.

If anyone can have a professional input and a solution to daemonise that would be greatly appreciated.

Regards
Mike

Code:

#!/bin/bash

rm -rf temp.mp3
rec temp.mp3 silence 1 5 4% 1 0:00:01 4%


for i in temp.mp3 ; do
   b=`stat -c %s "$i"`
if [ $b -ge 800 ] ; then

NAME=`date +%Y-%m-%d_%H-%M-%S`
TIME=`date +%H:%M:%S`
FILENAME=/var/www/Recordings/$NAME.mp3
FILEWWW=Recordings/$NAME.mp3
mv temp.mp3 $FILENAME
rm -rf temp.mp3

mysql --host=localhost --user=root --password=password database << EOF
insert into recordings (id,time,filename,active,status) values('NULL','$TIME','$FILEWWW','1','1');
EOF

~/vox



else
rm -rf temp.mp3
echo 'no sound'
fi
done

Freddythunder · 07-19-2014, 08:39 PM

I stumbled upon this post while looking for a solution to something I'm working on.. Don't know if it could help you now, but it looks like sox has a built in function that will split up a file by silences and put them into separate files. Here's a site I found that mentions it... I have not tried it
http://digitalcardboard.com/blog/200...ox-of-silence/

Maybe you can suggest something for me? I'd like to have my computer listen to a webstream and save segments of audio between silences. I was planning on trying to feed sox a URL for the input but I'm pretty sure that won't work. Thansk

LADave · 06-03-2015, 02:48 AM

And does SOX have a way to take input and do the VOX thing starting a new file each time etc. Meanwhile take the same input and replicate what a sound meter does, so you get a second file giving you a series of sound levels (with no breaks)? A sound meter would give you readings every 35 ms (impulse), 125 ms (fast) or 1 s (slow) with "A", "C", flat, or some other weighting scheme. Perhaps even produce more than one file using different weightings and time periods, maximum vs. RMS, etc.

duncanlinux · 09-25-2018, 08:57 AM

I know this is an old thread, but it seems like you could have this script run every minute using CRON then exit upon completion. The only problem would be you would be if you had a recording that lasted 15 seconds, you would have to wait 45 seconds before CRON would run again. People more versed in CRON than me may have a solution to run it in smaller intervals than 60 seconds.