21

I am looking for a tool that will tell me, in less than half a second, if the microphone is picking up any sound above a certain threshold. (I plan to then mute the Master channel with another command line tool, like amixer.)

Franck Dernoncourt
  • 20,384
  • 48
  • 186
  • 322
Christian Neverdal
  • 662
  • 1
  • 5
  • 21

4 Answers4

10

This solution will avoid writing repeatedly to disk, and even though it in worst case takes a second instead of the desired less than half a second, I found it to be fast enough after trying it. So, here are the two scripts I use:

./detect:

while true; do
    arecord -d 1 /dev/shm/tmp_rec.wav ; sox -t .wav /dev/shm/tmp_rec.wav -n stat 2>\
    &1 | grep "Maximum amplitude" | cut -d ':' -f 2 | ./check.py
    if [ $? -eq 0 ] ; then
         amixer set Master 0
    else
         amixer set Master 80
    fi
done

./check.py:

#!/usr/bin/env python
import sys

number = 0.0
thing="NO"

line = sys.stdin.readline()
thing = line.strip()
number = float(thing)

if number < 0.15:
    raise Exception,"Below threshold"

Hardly elegant, but it works.

Note: If you want a more gradual thing, add something like this:

   for i in `seq 0 80 | tac`; do
      amixer set Master $i
   done

for muting and

   for i in `seq 0 80`; do
      amixer set Master $i
   done

for unmuting.

Christian Neverdal
  • 662
  • 1
  • 5
  • 21
  • 8
    A slightly more elegant solution, which supports half-second resolution and doesn't require a temporary file: `while true; do amixer set Master $(rec -n stat trim 0 .5 2>&1 | awk '/^Maximum amplitude/ { print $3 < .15 ? 80 : 0 }'); done` – nandhp Jun 29 '12 at 13:17
  • 1
    Python is a bit overkill, http://math-blog.com/2012/07/23/floating-point-arithmetic-in-the-bourne-again-shell-bash/ result=$(AUDIODEV=hw:1 rec -n stat trim 0 .5 2>&1 | grep "Maximum amplitude" | grep -o "[0-9]\.[0-9]*$"); echo "$result > 0.01" | bc – Kevin Jun 15 '16 at 05:55
  • 1
    Just keep in mind that 'Maximum amplitude' isn't the only indicator of a loud sound. A sound with a high frequency (e.g. clinking of glasses) may be perceived as really loud by human ears but the sox' 'Maximum amplitude' won't be very different from the lowest one. So in some cases it would make sense to analyze 'Rough frequency' as well. – ka3ak May 19 '17 at 15:05
  • In contrast to `arecord`, I could not use `rec` while the input device is used by another process (webRTC using uv4l and janus webrtc). @Kevin @nandhp – Christopher K. May 31 '21 at 16:44
2

Just version without python script and TALKING_PERIOD, that sets up how many seconds will sound be on DOWN_SOUND_PERC level, then goes to UP_SOUND_PERC level.

#!/bin/bash
    
TALKING_PERIOD=16
UP_SOUND_PERC=65
DOWN_SOUND_PERC=45
counter=0

while true; do
    echo "counter: " $counter
    if [ "$counter" -eq 0 ]; then
        nmb=$(arecord -d 1 /dev/shm/tmp_rec.wav ; sox -t .wav /dev/shm/tmp_rec.wav -n stat 2>&1 | grep "Maximum amplitude" | cut -d ':' -f 2)

        echo "nmb: " $nmb

        if (( $(echo "$nmb > 0.3" |bc -l) )); then
            echo "ticho"
            amixer -D pulse sset Master 45%
            counter=$TALKING_PERIOD
        else
            echo "hlasno"
            amixer -D pulse sset Master 65%
        fi
    fi

    if [[ $counter -gt 0 ]]; then
        ((counter--))
    fi

    sleep 1

done
R4v0
  • 21
  • 1
0

There is a tool called pavumeter that lets you see the microphone level, Open capture interface of pavumeter,

Then adjust the capture sound level using pavucontrol, In pavucontrol, go to input devices, and adjust microphone sensitivity.

Edit: In the bash script by R4v0, done is inside code.

Edit2: I wanted to raise the volume each time there is noise, so i just edited more than to be less than and cancelled talking peroid

    if (( $(echo "$nmb < 0.3" |bc -l) )); then
0

I edited it to work with ffmpeg

sox and bc need to be installed.

#!/bin/bash

ffmpeg -y -loglevel panic -f pulse -i alsa_input.pci-0000_00_1b.0.analog-stereo -t 0.5 /dev/shm/tmp_rec.wav ; sox -t .wav /dev/shm/tmp_rec.wav -n stat 2>&1 | grep "Maximum amplitude" | cut -d ':' -f 2
  • And set end-of-line in your code editor to unix. – haytham-med haytham Aug 31 '19 at 09:16
  • Could I use this to 1. listen for a noise level, 2. 'fade' in playing an audio file, 3. timeout after x minutes, 4. 'fade' out and stop audio, 5. if noise level met again resume from previous point in said audio file? The fade in fade out functionality isn't as important as the resume portion for me. Would this be easy to implement? I'm trying to figure out how to make an automated white noise generator using bash and standard utilities if possible. I almost made my own post but this seems part way to the solution I'm looking for. – 66tree Sep 07 '19 at 16:50