openmmla 0.1.0

OpenMMLA: A unified multimedia platform integrating audio and vision modules for analysis and computer aided instruction.

OpenMMLA

OpenMMLA a toolkit for multimodal learning analytics, providing various built-in pipelines for different tasks. The toolkit is for building up the MMLA pipeline as shown below:

High-level System Design (Data flow)

The platform's high-level design consists of three stages: input, processing, and output.

• Data input stage (purple), multimodal raw data from sensors & wearable badges are streamed directly to base stations or a central media server. The raw inputs are transformed into structured, coded streams for efficient transmission and processing.
• Data processing stage (black), these encoded streams are processed individually by the corresponding (Audio/Video/..)Base, which handles some signal processing, while more complex tasks are offloaded to the server. For Bases within the same group, results are synchronized and uploaded to the time series database, where segment-level measurement features are generated.
• Data output stage (lime), these measurement features are visualized on the dashboard in real time and combined into indicators to analyze group interactions. The platform also generates post-processing visualizations, logs, and reports, which are stored and accessible via the shared dashboard, enabling both real-time awareness and retrospective analysis of group dynamics.

System Architecture

OpenMMLA consists of several hardware components working together:

• Sensors: wearable devices and distributed environmental sensors for data acquisition. As for wearable devices, we support:
  • Voice-Badge: Nicla Vision board with power supply
  • Vision-Badge: Nicla Vision board with AprilTag and power supply
  • Regular-Badge: AprilTag only
• Base Stations: microprocessors/PCs that processes various data streams. Each base station runs one or more instances of Base and Synchronizer, with specific types (e.g., AudioSynchronizer) synchronizing data from corresponding Base components (e.g., AudioBase).
• Servers: powerful PCs that provides centralized services for other devices within distributed environments. Based on functionality, it can be divided into:
  • Base Servers: REST servers for AI services (infer, transcribe, vad, vlm... via Flask/FastAPI).
  • Uber Servers: other services like database (InfluxDB), Messaging (Redis, MQTT), RTMP streaming & Load balancing (Nginx), dashboard application (Next.js & Flask).
• Dashboard: web-page interfaces accessible via phone and web browsers, featuring on session selection, real-time visualizations, post-time visualizations and measurements downloads.

Detailed Architecture

Quick Setup

The setup requirements depend on which part of the system you're implementing:

System Dependencies

These tools are required for Base Stations and Base Servers to process audio/video data:

• Python Environment: Conda
• Terminal Multiplexer: tmux
• Audio Processing: PortAudio
• Video Processing: FFmpeg

Conda Installation

# macOS & Ubuntu
wget "https://repo.anaconda.com/miniconda/Miniconda3-latest-$(uname)-$(uname -m).sh"
bash Miniconda3-latest-$(uname)-$(uname -m).sh

# Debian (Raspberry Pi)
wget "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
bash Miniforge3-$(uname)-$(uname -m).sh

Others Installation

# macOS
brew install ffmpeg portaudio mecab llvm tmux
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
sudo apt update && sudo apt upgrade
sudo apt install build-essential git ffmpeg python3-pyaudio libsndfile1 libasound-dev tmux
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

Debian (Raspberry Pi Bullseye or later)

sudo apt update && sudo apt upgrade
sudo apt install -y build-essential git ffmpeg python3-pyaudio libsndfile1 portaudio19-dev tmux

Required Services

These services are primarily required for the Uber Server which acts as the central hub:

• InfluxDB: Time series database for storing measurements data(Required)
• Redis: Message broker and cache (Required)
• Mosquitto: MQTT broker for publish/subscribe messaging (Required)
• Nginx: RTMP server and load balancer (Optional)
• Dashbaord NEXT.js frontend & Flask backend server (Optional)

Services Installation

InfluxDB Installation

# For macOS
brew install influxdb
brew services start influxdb

# For Ubuntu/Debian
wget -q https://repos.influxdata.com/influxdata-archive_compat.key
echo '393e8779c89ac8d958f81f942f9ad7fb82a25e133faddaf92e15b16e6ac9ce4c influxdata-archive_compat.key' | sha256sum -c && cat influxdata-archive_compat.key | gpg --dearmor | sudo tee /etc/apt/trusted.gpg.d/influxdata-archive_compat.gpg > /dev/null
echo 'deb [signed-by=/etc/apt/trusted.gpg.d/influxdata-archive_compat.gpg] https://repos.influxdata.com/debian stable main' | sudo tee /etc/apt/sources.list.d/influxdata.list
sudo apt update && sudo apt install influxdb2
sudo systemctl enable influxdb
sudo systemctl start influxdb

# Go to http://localhost:8086, and follow the instructions to create admin user with operator API token, save your token in a safe place, 
# it will be used for config later.

Redis Installation

# For macOS
brew install redis
# Edit /opt/homebrew/etc/redis.conf: set "protected-mode no" and "bind 0.0.0.0"
brew services restart redis

# For Ubuntu/Debian
sudo apt install redis-server
# Edit /etc/redis/redis.conf: set "protected-mode no" and "bind 0.0.0.0"
sudo systemctl enable redis-server
sudo systemctl restart redis-server

Mosquitto Installation

# For macOS
brew install mosquitto
# Edit /opt/homebrew/etc/mosquitto/mosquitto.conf: add "listener 1883 0.0.0.0" and "allow_anonymous true"
brew services restart mosquitto

# For Ubuntu/Debian
sudo apt install -y mosquitto
# Edit /etc/mosquitto/mosquitto.conf: add "listener 1883 0.0.0.0" and "allow_anonymous true"
sudo systemctl enable mosquitto
sudo systemctl restart mosquitto

Nginx Installation (Optional)

For detailed Nginx installation and configuration as both a load balancer and RTMP server, please refer to the Nginx Setup Guide.

Dashboard Installation (Optional)

For detailed instructions on setting up the Next.js frontend and Flask backend for the dashboard, please refer to the Dashboard Setup Guide.

OpenMMLA Installation

You can install OpenMMLA in two ways:

Option 1: Via PyPI

# Create conda environment
conda create -n openmmla python=3.10.12 -y
conda activate <your-env-name> e.g., asr-base, asr-server, ips-base, vfa-server

# Install base package
pip install openmmla

# Install pipeline-specific dependencies for specific envs
pip install 'openmmla[asr-base]'  # For ASR Base
pip install 'openmmla[asr-server]'  # For ASR Server
pip install 'openmmla[ips-base]'  # For IPS Base
pip install 'openmmla[vfa-server]'  # For VFA Server

Option 2: Via Git

# Clone the repository
git clone https://github.com/ucph-ccs/openmmla.git
cd openmmla

# Create conda environment
conda create -n openmmla python=3.10.12 -y
conda activate <your-env-name> e.g., asr-base, asr-server, ips-base, vfa-server

# Install in development mode
pip install -e .  # Install base package

# Install pipeline-specific dependencies for specific envs
pip install -e '.[asr-base]'  # For ASR Base
pip install -e '.[asr-server]'  # For ASR Server
pip install -e '.[ips-base]'  # For IPS Base
pip install -e '.[vfa-server]'  # For VFA Server

Pipeline Setup

After installing OpenMMLA and its dependencies, you can set up specific pipelines:

1. Automatic Speech Recognition (ASR) with Diarization

   • See ASR Pipeline README for detailed setup and usage instructions.

2. Indoor Positioning System (IPS)

   • See IPS Pipeline README for detailed setup and usage instructions.

3. Video Frame Analyzer (VFA)

   • See VFA Pipeline README for detailed setup and usage instructions.

Each pipeline has specific usage instructions and configuration options detailed in their respective documentation.

FAQ

Citation

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

@inproceedings{10.1145/3706468.3706525,
author = {Li, Zaibei and Yamaguchi, Shunpei and Spikol, Daniel},
title = {OpenMMLA: an IoT-based Multimodal Data Collection Toolkit for Learning Analytics},
year = {2025},
doi = {10.1145/3706468.3706525},
booktitle = {Proceedings of the 15th International Learning Analytics and Knowledge Conference},
pages = {872–879},
}