openmmla-audio 0.1.4.post5

Audio module for the OpenMMLA platform

🎙️ OpenMMLA Audio

Audio module of the mBox - an open multimodal learning analytic platform. For more details, please refer to mBox System Design.

Table of Contents

• Related Modules
• Installation
  • Uber Server Setup
  • Audio Base & Server Setup
  • Standalone Setup
• Usage
  • Real-time Audio Analyzer
  • Post-time Audio Analyzer
• Log & Visualization
• FAQ
• Citation
• References
• License

Related Modules

• mbox-uber
• mbox-video

Installation

Uber Server Setup

Before setting up the audio base, you need to set up a server hosting the InfluxDB, Redis, and Mosquitto services. Please refer to mbox-uber module.

Audio Base & Server Setup

1. Clone the repository

   git clone https://github.com/ucph-ccs/mbox-audio.git
   

2. Install required system dependencies

   Mac

   brew install ffmpeg portaudio mecab llvm
   echo 'export PATH="/opt/homebrew/opt/llvm/bin:$PATH"' >> ~/.zshrc
   echo 'export LDFLAGS="-L/opt/homebrew/opt/llvm/lib"' >> ~/.zshrc
   echo 'export CPPFLAGS="-I/opt/homebrew/opt/llvm/include"' >> ~/.zshrc
   source ~/.zshrc
   

   Ubuntu 24.04

   sudo apt update && sudo apt upgrade
   sudo apt install build-essential git ffmpeg python3-pyaudio libsndfile1 libasound-dev
   wget https://files.portaudio.com/archives/pa_stable_v190700_20210406.tgz
   tar -zxvf pa_stable_v190700_20210406.tgz
   cd portaudio
   ./configure && make
   sudo make install
   

   Raspberry Pi Bullseye or later

   sudo apt-get install portaudio19-dev
   

3. Install openmmla-audio

   Set up Conda environment

   # For Raspberry Pi
   wget "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
   bash Miniforge3-$(uname)-$(uname -m).sh
   
   # For Mac and Linux
   wget "https://repo.anaconda.com/miniconda/Miniconda3-latest-$(uname)-$(uname -m).sh"
   bash Miniconda3-latest-$(uname)-$(uname -m).sh
   

   Install Audio Base

   conda create -c conda-forge -n audio-base python==3.10.12 -y
   conda activate audio-base
   pip install openmmla-audio[base]  # for Linux and Raspberry Pi
   pip install 'openmmla-audio[base]'  # for Mac
   

   Install Audio Server

   conda create -c conda-forge -n audio-server python==3.10.12 -y
   conda activate audio-server
   pip install openmmla-audio[server]  # for Linux and Raspberry Pi
   pip install 'openmmla-audio[server]'  # for Mac
   

4. Set up folder structure

   cd mbox-audio
   ./reset.sh
   

Standalone Setup

If you want to run the entire mBox Audio system on a single machine (not in distributed mode), follow these steps:

1. Set up the Uber Server on your machine following the instructions in the mbox-uber module.

2. Install system dependencies as described in the "Audio Base & Server Setup" section above.

3. Install openmmla-audio with all dependencies:

   conda create -c conda-forge -n mbox-audio python==3.10.12 -y
   conda activate mbox-audio
   pip install openmmla-audio[all]  # for Linux and Raspberry Pi
   pip install 'openmmla-audio[all]'  # for Mac
   

4. Set up the folder structure:

   cd mbox-audio
   ./reset.sh
   

This setup will allow you to run all components of mBox Audio on a single machine.

Usage

Real-time Audio Analyzer

To run the real-time audio analyzer:

1. Start Audio Server (optional)

   ./server.sh
   

   This script runs distributed audio services on audio servers. To configure your audio server cluster: please refer to the nginx setup running on your uber server.Default setup: Three servers (server-01.local, server-02.local, server-03.local) with five services (transcribe, separate, infer, enhance, vad).

2. Start Audio Base

   ./run.sh [-b <num_bases>] [-s <num_synchronizers>] [-l <standalone>] [-p <speech_separation>]
   

   Default parameter settings:

   • -b: 3 (number of audio base)
   • -s: 1 (number of audio base synchronizer)
   • -l: false (not standalone)
   • -p: false (no speech separation)

   💡 You can switch the operating mode of the audio base during runtime:

   Mode	Description
   Record	Record audio segments without recognition
   Recognize	Recognize pre-recorded segments and synchronize
   Full	Default: record and recognize simultaneously

3. Start Control Base

   ./control.sh
   

Post-time Audio Analyzer

To run the post-time audio analyzer:

1. Create a speaker corpus folder: /audio_db/post-time/[audio_file_name]/
2. Add speaker audio files named [speaker_name].wav to the corpus folder
3. Run the analyzer:

   cd examples/
   conda activate mbox-audio
   
   # process single audio file (supported formats: wav, m4a, mp3), if not specified, then would process all files in the /audio/post-time/origin/ folder
   python3 run_audio_post_analyzer.py -f [audio_file_name.wav] 
   

Logs & Visualization

After running the analyzers, logs and visualizations are stored in the /logs/ and /visualizations/ folders.

FAQ

Citation

If you use this code in your research, please cite the following paper:

@inproceedings{inproceedings,
  author = {Li, Zaibei and Jensen, Martin and Nolte, Alexander and Spikol, Daniel},
  year = {2024},
  month = {03},
  pages = {785-791},
  title = {Field report for Platform mBox: Designing an Open MMLA Platform},
  doi = {10.1145/3636555.3636872}
}

References

• NeMo
• Silero VAD
• Denoiser
• MossFormer
• Whisper

License

This project is licensed under the MIT License - see the LICENSE file for details.