Hiss / White / Static noise cancellation on Linux using Pulseaudio and Sox

noise-cancellation.sh

#!/bin/bash

# You'll need to have sox, pavucontrol and alsa-utils installed, and the snd_aloop kernel module loaded. 
# You can configure your system to load it on startup or load it manually with "sudo modprobe snd_aloop"

# Once this is script is running, you need to start recording audio in the application of your
# preference, open pavucontrol, go to the recording tab and change the recording source of that application 
# to "Monitor of Loopback ..."

time=5
workDir='/tmp'
sampleRate=44100

# get pulse audio devices
devices=`pactl list | grep -E -A2 '(Source|Sink) #' | grep 'Name: ' | grep -v monitor | cut -d" " -f2`
if [ `echo "$devices" | grep -c aloop` -lt 1 ]; then
    echo "No loopback device created. Run 'sudo modprobe snd_aloop' first."
    exit
fi
 
input=`echo "$devices" | grep input.*pci`
output=`echo "$devices" | grep output.*aloop`

record()
{
    echo "Recording background noise. Keep quiet for $time seconds."
    sleep 3
    pacat -r -d $input --latency=1msec | sox -b 16 -c 2 -e signed -t raw -r $sampleRate - -b 16 -c 2 -e signed -r $sampleRate -t wav noise.wav &
    PID=$!
    sleep $time
    kill $PID
    aplay noise.wav
}
  
cd $workDir

if [ ! -f "noise.prof" ] || [ "$1" == "--record" ]; then
    # record noise sample and create noise profile
    record
    sox noise.wav -n noiseprof noise.prof
fi

GREEN='\033[0;32m'
WHITE='\033[1;37m'
NC='\033[0m'

echo -e "

${GREEN}Redirecting the filtered output to $output...${NC}

    Start recording in the app of your preference, open pavucontrol, go to the recording tab and change the recoding source to \"Monitor of Loopback ...\".
    ${WHITE}You'll need to repeat this process for every application (chrome, skype, etc...)${NC}. But this is a persistent configuration, only need to be done once.
    
"
pacat -r -d $input --latency=1msec | sox -v 0.8 -b 16 -c 2 -e signed -t raw -r $sampleRate - -C 0.5 -b 16 -c 2 -e signed -r $sampleRate -t raw - noisered noise.prof 0.21 | pacat -p -d $output --latency=1msec

Thanks for the suggestions @pklapperich, I've updated it. Sorry for the late response but this is something I haven't used in a long time, and I don't really use it anymore.

I followed the process inside Ubuntu 18.04 and I couldn't find "Monitor of Loopback..." in the pavucontrol.

@DevotionGeo you need to add " around input.*pci and output.*aloop.
But I still wasn't able to get any meaningful results, like improvement in the recordings quality.

Thanks for the suggestions @pklapperich, I've updated it. Sorry for the late response but this is something I haven't used in a long time, and I don't really use it anymore.

did you have another method? @tiagoengel

Thanks for the suggestions @pklapperich, I've updated it. Sorry for the late response but this is something I haven't used in a long time, and I don't really use it anymore.

did you have another method? @tiagoengel

Not really, because of work I've been using a macbook for the past few years, so this is not an issue anymore.

How do I REMOVE the permanent configuration? I did
sudo apt-get remove --purge alsa pulseaudio sox
and then re-installed, but still my application (firefox) selects "Monitor of device" as input source, and when I select another source in pulseaudio it won't change it :(

How do I REMOVE the permanent configuration? I did
sudo apt-get remove --purge alsa pulseaudio sox
and then re-installed, but still my application (firefox) selects "Monitor of device" as input source, and when I select another source in pulseaudio it won't change it :(

I removed noise.prof and noise.wav from /tmp and rebooted. that worked for me!

How do I REMOVE the permanent configuration? I did
sudo apt-get remove --purge alsa pulseaudio sox
and then re-installed, but still my application (firefox) selects "Monitor of device" as input source, and when I select another source in pulseaudio it won't change it :(

I removed noise.prof and noise.wav from /tmp and rebooted. that worked for me!

Thanks for the tip, but unfortunately I don't have those files and it still does not work.
UPDATE: it seems like firefox settings was actually the culprit. I believe when you let services remember a specific shared microhpone, it remembers to a) never ask for permission again and b) to use only the microphone it has permissions for.
Must be a big coincidence then, because I never had that issue before I used this script.