odas_ros

This package is a ROS package for ODAS.

Authors

Marc-Antoine Maheux
Cédric Godin
Simon Michaud
Samuel Faucher
Olivier Roy
Vincent Pelletier
Philippe Warren
François Grondin

License

Prerequisites

You will need CMake, GCC and the following external libraries:

sudo apt-get install cmake gcc build-essential libfftw3-dev libconfig-dev libasound2-dev libpulse-dev libgfortran-*-dev perl

ODAS ROS uses the audio utilities from AudioUtils so it should be installed in your catkin workspace. If it is not already, here is how to do so:

Clone AudioUtils in your catkin workspace:

git clone https://github.com/introlab/audio_utils.git

Install dependencies:

sudo apt-get install gfortran texinfo
sudo pip install libconf

In the cloned directory of audio_utils, run this line to install all submodules:

git submodule update --init --recursive

If you get errors when building with catkin_make, you can modify the cmake file of audio_utils to add C++ 14 compiler option.

add_compile_options(-std=c++14)

Installation

First, you need to clone the repository in your catkin workspace.

git clone https://github.com/introlab/odas_ros.git

In the cloned directory of odas_ros, run this line to install all submodules:

git submodule update --init --recursive

Hardware configuration

For ODAS to locate and track sound sources, it needs to be configured. There is a file (configuration.cfg) that is used to provide ODAS with all the information it needs. You will need the position and direction of each microphones. See ODAS Configuration for details.

Here are the important steps:

Input configuration

Source configuration using pulseaudio

At this part of the configuration file, you need to set the correct pulseaudio device and channel mapping.

# Input with raw signal from microphones
    interface: {
        type = "pulseaudio";
        #"pacmd list-sources | grep 'name:' && pacmd list-sources | grep 'channel map:'" to see the sources and their channel mapping, in the same order
        source = "alsa_input.usb-SEEED_ReSpeaker_4_Mic_Array__UAC1.0_-00.multichannel-input";
        channelmap = ("front-left", "front-right", "rear-left", "rear-right", "front-center", "lfe");

To know your source name and channel mapping, the easiest way it to use pacmd list-sources | grep 'name:' && pacmd list-sources | grep 'channel map:' in a terminal. The output should look something like this:

	name: <alsa_output.pci-0000_00_1f.3.analog-stereo.monitor>
	name: <alsa_input.pci-0000_00_1f.3.analog-stereo>
	name: <alsa_input.usb-SEEED_ReSpeaker_4_Mic_Array__UAC1.0_-00.multichannel-input>
	channel map: front-left,front-right
	channel map: front-left,front-right
	channel map: front-left,front-right,rear-left,rear-right,front-center,lfe

Note that the names and channel maps are in the same order. In this case, the source name is alsa_input.usb-SEEED_ReSpeaker_4_Mic_Array__UAC1.0_-00.multichannel-input and the mapping is front-left,front-right,rear-left,rear-right,front-center,lfe. The mapping will need to be formated in a list: front-left,front-right,rear-left,rear-right,front-center,lfe will become channelmap = ("front-left", "front-right", "rear-left", "rear-right", "front-center", "lfe");

Alternative: sound card configuration using ALSA

At this part of the configuration file, you need to set the correct card and device number.

# Input with raw signal from microphones
    interface: {    #"arecord -l" OR "aplay --list-devices" to see the devices
        type = "soundcard_name";
        devicename = "hw:CARD=1,DEV=0";

To know what is your card number, plug it in your computer and run arecord -l in a terminal. The output should look something like this:

**** List of CAPTURE Hardware Devices ****
card 0: PCH [HDA Intel PCH], device 0: ALC294 Analog [ALC294 Analog]
  Subdevices: 1/1
  Subdevice #0: subdevice #0
card 1: 8_sounds_usb [16SoundsUSB Audio 2.0], device 0: USB Audio [USB Audio]
  Subdevices: 1/1
  Subdevice #0: subdevice #0

In this case, the card number is 1 and the device is 0 for the 16SoundsUSB audio card. So the device name should be: "hw:CARD=1,DEV=0";.

Mapping

Depending on your configuration, you will need to map the microphones from the soundcard to the software. If you wish to use all microphones, then you can map all of them. For example, if there is 16 microphones and wish to use them all:

mapping:
{
    map: (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16);
};

But if only certain microphones should be used, they can be mapped. For example, if I only wish to use microphones 1 to 4 and 6 to 10 and 12:

mapping:
{
    map: (1,2,3,4,6,7,8,9,10,12);
};

They will be mapped to microphones 1 to 10.

Microphone configuration

For ODAS to precisely locate and track a sound source, it needs to know precisely the microphone position. The frame of reference used to measure the microphones position will end up being the one used for the sound tracking. Here is an example of a microphone configuration. It is easier if the reference point is located in the center of the microphones.

# Microphone 1 #TODO
        {
            mu = ( 0.122376, 0.08144437, 0.042 );
            sigma2 = ( +1E-6, 0.0, 0.0, 0.0, +1E-6, 0.0, 0.0, 0.0, +1E-6 );
            direction = ( -0.08144, -0.12238, 0.0 );
            angle = ( 80.0, 100.0 );
        },

For Microphone 1, mu is the position in x, y and z from the reference point. sigma2 is the position variance in xx, xy xz, yx, yy, yz, zx, zy, zz this setting should mainly remain untouched. The direction parameter is the direction of the microphone. It should be a unit vector pointing in the direction that the microphone is pointing relative to the reference frame. The angle parameter is the maximum angle at which gain is 1 and minimum angle at which gain is 0.

Sound Source Localization, Tracking and Separation

ODAS can output the sound source localization, the source source tracking and the sound source separation:

Sound Source Localization: All the potential sound sources in the unit sphere. Each sound source position on the unit sphere and its energy.
Sound Source Tracking: The most probable location of the sound source is provided (xyz position on the unit sphere).
Sound Source Separation: An Audio Frame of the isolated sound source is provided.

Depending on what type of information will be used, the configuration file needs to be modified. For example, if I need only the Sound Source Tracking, the configuration file should be modified. The only thing that should be changed is the format and interface for each type of data. The required format if it is enabled is json and the interface type should be socket. If it is disabled, the format can be set to undefined and the interface type to blackhole.

For example, if the only data that will be used is the sound source tracking:

In the # Sound Source Localization section, this should be modified to look like this:

potential: {

        #format = "json";

        #interface: {
        #    type = "socket";
        #    ip = "127.0.0.1";
        #    port = 9002;
        #};

        format = "undefined";

        interface: {
           type = "blackhole";
        };

In the # Sound Source Tracking section, this should be modified to look like this:

# Output to export tracked sources
    tracked: {

        format = "json";

        interface: {
            type = "socket";
            ip = "127.0.0.1";
            port = 9000;
        };
    };

In the # Sound Source Separation section, this should be modified to look like this:

separated: { #packaging and destination of the separated files

        fS = 44100;
        hopSize = 256;
        nBits = 16;

        interface: {
           type = "blackhole";
        };

        #interface: {
        #    type = "socket";
        #    ip = "127.0.0.1";
        #    port = 9001;
        #}
    };

Note that if an interface type is set to "blackhole" and the format to "undefined", the associated topic won't be published.

Sound Source Tracking Threshold adjustment

The default configuration file should be correct for most configuration. However, if the Sound Source Tracking does not work (i.e. the published topic /odas/sst does not contain any sources or the sources are indesirable) it may be because the threshold is not set correctly.

In the Source Source Tracking section of the configuration file, there is a section with active and inactive:

# Parameters used by both the Kalman and particle filter

   active = (
       { weight = 1.0; mu = 0.3; sigma2 = 0.0025 }
   );

   inactive = (
       { weight = 1.0; mu = 0.15; sigma2 = 0.0025 }
   );

The active parameter represents the limit to consider a sound source active (high limit) and theinactive parameter is the lower limit at which a sound source is considered inactive.

If mu is too high in the active and inactive parameters, few sound sources will be considered like active.
If mu in the active and inactive parameters are set too low, too much sound sources will be considered active.

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

introlab / odas_ros

Programming Languages