Skip to content

Multiclass classification example/exercise using deep neural networks (DNNs)

License

Notifications You must be signed in to change notification settings

spatialaudio/classification_exercise

Repository files navigation

Classification of Bulk Material by Structure Borne Sound

This repository contains an example for multiclass classification as introduction to machine learning on audio signals. Different bulk materials create characteristic structure borne sounds when rolling/slipping down a ramp. Audio recordings from different types of screws and bolts rolling down an aluminium ramp have been conducted. A deep neural network (DNN) is trained and evaluated for the classification of the type of bulk metarial rolling down the ramp.

Getting Started

  1. Install a Python 3.7 environment on your computer, e.g. Anaconda, including support for jupyter notebooks
  2. Check if the following Python packages are installed
    • numpy
    • matplotlib
    • pysoundfile
    • tensorflow
    • keras
    • scikit-learn
  3. Clone the repository
  4. Open the jupyter notebook train_model.ipynb and run all cells
  5. Take a look at the exercises at the end of the notebook

License

The notebooks are provided as Open Educational Resources. Feel free to use the notebooks for your own purposes. The text/images/data are licensed under Creative Commons Attribution 4.0, the code of the IPython examples under the MIT license.