ROS audio_common Package

This repository is a fork from the ros_drivers/audio_common containing the ROS audio_common package and modifications for the AGC 2022 challenge.

Building from source

For the AGC 2022 competition, please git clone and build the feature/wave branch using catkin_make

cd ~/catkin_ws/src
git clone https://github.com/swl017/audio_common -b feature/wave
cd ..
rosdep install --from-paths src/ --ignore-src -r -y
catkin_make

Mic check

If arecord -l shows the SpkUAC20 on card 2, the mic array is set and good to go.

(Example)

$ arecord -l
**** List of CAPTURE Hardware Devices ****
card 1: Generic [HD-Audio Generic], device 0: ALCS1200A Analog [ALCS1200A Analog]
  Subdevices: 1/1
  Subdevice #0: subdevice #0
card 1: Generic [HD-Audio Generic], device 2: ALCS1200A Alt Analog [ALCS1200A Alt Analog]
  Subdevices: 1/1
  Subdevice #0: subdevice #0
card 2: SpkUAC20 [miniDSP VocalFusion Spk (UAC2.0], device 0: USB Audio [USB Audio]
  Subdevices: 0/1
  Subdevice #0: subdevice #0

The concept and usage

이 패키지의 컨셉은 젯슨(audio_pub.py)에서 8 채널 오디오 데이터를 /audio/all/ 토픽으로 publish하고, 서버(audio_sub.py)에서 subscribe 받은 메세지를 각 채널로 분리하여 사용하는 것입니다. 따라서 아래와 같이 실행합니다.
The concept is to publish 8 channel audio data to the topic /audio/all from Jetson(audio_pub.py), subscribe to the topic and split the audio channels on the server(audio_sub.py). Thus, the nodes are run as the following:

1. Jetson side

rosrun audio_capture audio_pub.py

This publishes audio data including all 8 channels in audio/all topic.
The parameters are published in audio/all/info topic and are used in the subscriber node described below.
For the moment, only pyhon2 is supported and not python3 due to differences in writing data into bytes. Please push a PR if you are willing to share your solution.

2. Server side

rosrun audio_capture audio_sub.py

audio_sub.py subscribes to audio/all topic for the actual data and audio/all/info topic for referencing the format of the data.
The subscriber node seperates the 8 channel data into individual channels. While the individual channels are published in audio/ch{CHANNEL} topics, however, you may want to work with the seperated channels directly in the subscriber node to achieve maximum performance. (To get over with network delay, bandwidth, sync issues, etc.)
You must select an appropriate length for each audio file in file_length_sec variable.
The subscriber node saves the original 8 channel audio into wave_{NUMBER_CHANNELS}_{SAMPLE_FORMAT}_{SAMPLE_RATE}_{ROS_TIME_STAMP}.wav and the individual channels in wave_{NUMBER_CHANNELS}_{SAMPLE_FORMAT}_{SAMPLE_RATE}_{ROS_TIME_STAMP}_{CHANNEL}.wav. The audio parameters and the names come from audio/all/info topic if it is ever subscribed, else default values are used. File saving is tested valid on python2. Saving to .wav is not supported on python3 for now. Please push a PR if you are willing to share your solution.

Another possibility but not tested

capture_wave.launch from the original audio_common repository. Currently, it looks like mic arrays with channels more than 2 are not supported, though we need more investigation on this.

Further Reference

The purpose of the provided code is to demonstrate an example to handle multi-channel audio data in ROS environment. You can to modify the provided code to fit your own needs.