audio-segmentation-by-classification-tutorial

A jupyter notebook for a multiclass audio segmentation tutorial and demo.

What is it all about?

Whe have an audio stream and we want to automatically figure out, given some training data, which kinds of sound envents it contains and when each of these events occurs (see next figure).

You can read a static version of this notebook here.

Installation

Installation with Anaconda and conda

• Install Anaconda

• Clone this repository and run Jupyter:

    git clone https://github.com/amsehili/audio-segmentation-by-classification-tutorial.git
    cd audio-segmentation-by-classification-tutorial
    conda env create -f linux-64-environment.yml
    source activate audio
    jupyter notebook
  

Manual installation (using pip and virtualenv)

Clone the repository

git clone https://github.com/amsehili/audio-segmentation-by-classification-tutorial.git
cd audio-segmentation-by-classification-tutorial

Create a virtual environment (optional)

pip install virtualenv
virtualenv audio
source audio/bin/activate

Install requirements

The following packages are required:

• numpy
• scikit-learn
• matplotlib
• auditok
• librosa
• Jupyter

To install them all type:

pip install -r requirements.txt

If you have some of the previous packages already installed, you can install missing one by typing:

pip install package_name

Optionally, you can install pyaudio to play audio segments within the notebook. On Ubuntu type the following command to install the library:

$ sudo apt-get install python-pyaudio portaudio19-dev

Please refer to pyaudio web page for installation instructions on others platforms.

Run the notebook

jupyter notebook