egonoise

The goal of this project is to create a system able to reduce easily the egonoise of a robot using the Minimum Variance Distortionless Response (MVDR) algorithm. This branch is the new version of the project. The data is process faster. Last version is available on the other branch.

Author(s): Pierre-Olivier Lagacé

Installation

ROS Noetic on Ubuntu 20.04

1. Follow the instructions on the offical website (ROS Installation)
2. If you are not familiar with ROS, we strongly recommend that you do the tutorials (ROS Tutorials)

Configuring your Catkin Workspace and Installation

1. Make shure that the end of your .bashrc file in your home folder has the following lines

source /opt/ros/noetic/setup.bash
source ~/catkin_ws/devel/setup.bash

Python

1. Install Python 3.10.5
2. Clone this repo in your catkin_ws/src
3. Install the requirement.txt

pip3 install -r requirement.txt

4. Install kissdsp from https://github.com/FrancoisGrondin/kissdsp on your computer (not in the catkin_ws).

ROS Libraries

1. Install https://github.com/introlab/audio_utils in your catkin_ws/src

calibration_run.py

Expication: This node allow to train de database with a rosbag using the command roslaunch egonoise egonoise.launch calibration_run:=true.

Parameters:

• input_format
• database_path
• bag_calibration
• bag_calibration_path
• sampling_frequency
• frame_size
• frame_sample_count
• hop_length
• overlap
• calibration_step
• n_frame_scm
• n_batch

calibration_node.py

Expication: This node allow to train de database with live input using the command roslaunch egonoise egonoise.launch calibration_node:=true. Parameters:

• input_format
• database_path
• sampling_frequency
• frame_sample_count
• frame_size
• channel_count
• overlap
• hop_length
• calibration_duration
• step
• n_frame_scm

Topics (Sub and Pub)

• Sub: audio_out

egonoise_node.py

Expication: This node allow to use the framework to filtered the signal from subscriber using the command roslaunch egonoise egonoise.launch egonoise_node:=true

Parameters:

• input_format
• output_format
• database_path
• frame_size
• sampling_frequency
• channel_count
• overlap
• hop_length
• n_frame_scm
• n_batch

Topics (Sub and Pub)

• Sub: audio_out
• Pub: audio_in

Setup RaspberryPi

Info

username: ubuntu password: egonoise

Installation

1. Flash SD card with ubuntu 20.04 server 64bits using RaspberryPi imager
2. Launch RaspberryPi with SD card
3. Setup Wifi -> https://linuxconfig.org/ubuntu-20-04-connect-to-wifi-from-command-line
4. sudo apt-get update (and upgrade?)
5. Install Python: Pyenv -> https://k0nze.dev/posts/install-pyenv-venv-vscode
6. Tests with microphones array:

• Install pulse_audio utils
• if pactl list do pa_context_connect() failed: Connection refused try sudo apt-get --purge --reinstall install pulseaudio
• sudo apt-get install libportaudio2
• Test a python script to make a test record

8. Install Ros Noetic

• Use catkin_make -DPYTHON_EXECUTABLE=/usr/bin/python3 for the first catkin_make

9. Install KissDsp
10. Install audio_utils
11. Install this project in catkin_ws/src/
12. Test with: roslaunch egonoise egonoise.launch audio_capture:=true with the good parameter.
13. Follow the guide https://husarion.com/tutorials/ros-tutorials/5-running-ros-on-multiple-machines/ if you want to record the rosbag on another machine.