audio_utils

ROS2 nodes and utilities for audio streams.

For ROS1, please see the ros1 branch.

Author(s): Marc-Antoine Maheux

Setup (Ubuntu)

The following subsections explain how to use the library on Ubuntu.

Install Dependencies

bash sudo apt-get install cmake build-essential gfortran texinfo libasound2-dev libpulse-dev libgfortran-*-dev

Install Python Dependencies

bash sudo pip install -r requirements.txt

or

bash sudo pip3 install -r requirements.txt

Setup Submodules

bash git submodule update --init --recursive

Nodes

capture_node

This node captures the sound from an ALSA or PulseAudio device and publishes it to a topic.

Parameters

• backend (string): The backend to use (alsa or pulse_audio). The default value is alsa.
• device (string): The device to capture (ex: hw:CARD=1,DEV=0 or default for ALSA, or alsa_input.usb-IntRoLab_16SoundsUSB_Audio_2.0-00.multichannel-input for PulseAudio). The default value is default.
• format (string): The audio format ( see audioutilsmsgs/AudioFrame). The default value is signed_16.
• channel_count (int): The device channel count. The default value is 1.
• sampling_frequency (int): The device sampling frequency. The default value is 16000.
• frame_sample_count (int): The number of samples in each frame. The default value is 1024.
• merge (bool): Indicate to merge the channels or not. The default value is false.
• gain (double): The gain to apply. The default value is 1.0.
• latency_us (int): The capture latency in microseconds. The default value is 64000.
• channel_map (Array of string): The PulseAudio channel mapping. If empty or omitted, the default mapping is used. This parameter must be set only with the PulseAudio backend. The default value is [].
• queue_size (int): The publisher queue size. The default value is 1.

Published Topics

• audio_out (audioutilsmsgs/AudioFrame) The captured sound.

playback_node

This node captures the sound from a topic and plays it to an ALSA or PulseAudio device.

Parameters

• backend (string): The backend to use (alsa or pulse_audio). The default value is alsa.
• device (string): The device to capture (ex: hw:CARD=1,DEV=0 or default for ALSA, or alsa_input.usb-IntRoLab_16SoundsUSB_Audio_2.0-00.multichannel-input for PulseAudio). The default value is default.
• format (string): The audio format ( see audioutilsmsgs/AudioFrame). The default value is signed_16.
• channel_count (int): The device channel count. The default value is 1.
• sampling_frequency (int): The device sampling frequency. The default value is 16000.
• frame_sample_count (int): The number of samples in each frame. The default value is 1024.
• latency_us (int): The capture latency in microseconds. The default value is 64000.
• channel_map (Array of string): The PulseAudio channel mapping. If empty or omitted, the default mapping is used. This parameter must be set only with the PulseAudio backend. The default value is [].
• queue_size (int): The publisher queue size. The default value is 1.

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound to play.

beat_detector_node

This node estimates the song tempo and detects if the beat is in the current frame.

Parameters

• sampling_frequency (int): The device sampling frequency. The default value is 44100.
• frame_sample_count (int): The number of samples in each analysed frame. It must be a multiple of oss_fft_window_size. The default value is 128.
• oss_fft_window_size (int): The onset strength signal window size. It must be greater than or equal to frame_sample_count. The default value is 1024.
• flux_hamming_size (int): The flux hamming window size to calculate the onset strength signal. The default value is 15.
• oss_bpm_window_size (int): The onset strength signal window size to calculate the BPM value. The default value is 1024.
• min_bpm (double): The minimum valid BPM value. The default value is 50.
• max_bpm (double): The maximum valid BPM value. The default value is 180.
• bpm_candidate_count (int): The number of cross-correlations to perform to find the best BPM. The default value is 10.

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound to analyze. The channel count must be 1.

Published Topics

• bpm (std_msgs/Float32): The tempo in bpm (beats per minute) for each frame.
• beat (std_msgs/Bool): Indicate if the beat is in the current frame.

vad_node

This node performs voice activity detection with Silero VAD. The models folder contains the model trained by Silero VAD. The license of the model is MIT.

Parameters

• silence_to_voice_threshold (double): The threshold to detect voice activity when silence was previously detected. The default value is 0.5.
• voice_to_silence_threshold (double): The threshold to detect silence when voice activity was previously detected. It must be lower than silence_to_voice_threshold. The default value is 0.4.
• min_silence_duration_ms (double): The minimum silence duration in ms. The default value is 500.

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound to analyze. The channel count must be 1. The samply frequency must be 16000 Hz. The frame sample count must be a multiple of 512.

Published Topics

• voice_activity (audioutilsmsgs/VoiceActivity) The voice activity detection result.

format_conversion_node.py

This node converts the format of an audio topic.

Parameters

• input_format (string): The input audio format ( see audioutilsmsgs/AudioFrame).
• output_format (string): The output audio format ( see audioutilsmsgs/AudioFrame).

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound topic to convert.

Published Topics

• audio_out (audioutilsmsgs/AudioFrame) The converted sound.

resampling_node.py

This node resamples an audio topic.

Parameters

• input_format (string): The input audio format ( see audioutilsmsgs/AudioFrame).
• output_format (string): The output audio format ( see audioutilsmsgs/AudioFrame).
• channel_count (int): The device channel count.
• input_sampling_frequency (int): The input sampling frequency.
• output_sampling_frequency (int): The output sampling frequency.
• input_frame_sample_count (int): The number of samples in each frame of the input.
• dynamic_input_resampling (bool: default is false): If true, always adjust the input sampling informations ( format, sampling frequency and frame sample count) to the sampling informations of the reveiced frames, dynamically. In this mode, input_format, input_sampling_frequency and input_frame_sample_count are not required, but they can be used to save a recomputation if the starting input sampling informations are known.

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound topic to resample.

Published Topics

• audio_out (audioutilsmsgs/AudioFrame) The resampled sound.

split_channel_node.py

This node split a multichannel audio topic into several mono audio topics.

Parameters

• input_format (string): The input audio format ( see audioutilsmsgs/AudioFrame).
• output_format (string): The output audio format ( see audioutilsmsgs/AudioFrame).
• channel_count (int): The device channel count.

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound topic to split.

Published Topics

• audio_out_0 (audioutilsmsgs/AudioFrame) The first channel sound.
• audio_out_1 (audioutilsmsgs/AudioFrame) The second channel sound.
• ...

raw_file_writer_node.py

This node writes the raw sound data to a file.

Parameters

• output_path (string): The output file path.

Subscribed Topics

• audio_in (audioutilsmsgs/AudioFrame) The sound topic to write.

License

• GPL-3.0 License

Sponsor

IntRoLab - Intelligent / Interactive / Integrated / Interdisciplinary Robot Lab