Toolbox for unsupervised audio signal learning. Currently on development, only few tools are available.

First install anaconda, example use miniconda (ligth version of anaconda).

Next, create the environment with the .yml file of this project.

    conda env create -f conda_environment.yml

Don't forget to activate it using this project.

• python3
• numpy
• scipy
• librosa (audio reading and extracting features)
• matplotlib
• scikit-learn (machine learning algorithms)

python activity-detector/learn.py [-h] [-ml MAX_LEARN] [-v] [-l LOGFILE]
               folder_audio_in folder_out config

config is a JSON file, it contains model, feature and other details. Look at config/example_config.json for an example.

Type python activity-detector/learn.py -h for details.

python activity-detector/forward.py [-h] [-v] [-l LOGFILE]
                  folder_audio_in folder_model folder_out

Type python activity-detector/forward.py -h for details.

python activity-detector/plot_forwarded.py [-h] [-v] [-l LOGFILE]
                         folder_forwarded folder_out max_frequency

Type python activity-detector/plot_forwarded.py -h for details.

  python activity-detector/plotting_clusters/plot_informations.py <audio filename>