dtwmetrics 0.1.2

Dynamic time warping metrics.

dtw_metrics

Dynamic time warping metrics.

Description

Dynamic time warping (DTW) is an algorithm used to measure similarity between different data series, which may vary in speed.

Setup

Makefile

Setup via Makefile:

• Clone the repository first.

• Help and overview of commands:

  make help
  

• Setup (Create the venv, install libraries and pull minimal data setup):

  make setup
  

• Activate Python venv:

  make venv
  

• Clean project:

  make clean
  

Python venv (Poetry)

• Poetry is used for the virtual Python environment.

  • Install python poetry:

    curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python
    

    or

    pip install --user poetry
    

  • Make sure that the virtual environment is installed in the root folder of this projects:

    poetry config virtualenvs.in-project true
    

  • Install dependencies:

    poetry install --no-root
    

  • Add packages: To add further packages, run:

    poetry add <package-name>
    

Code formatting

• Formatting via pre-commit hook: Python code is formatted using pre-commit hooks.

• Install pre-commit:

  pip install pre-commit
  

  or

  brew install pre-commit
  

• Install pre-commit hooks from the config file .pre-commit-config.yaml

  pre-commit install
  

• Run the pre-commit hooks:

  pre-commit run -a
  

Code testing

Testing is done with pytest. The pytest package is already installed in the poetry venv.

• Run all tests:

  pytest
  

• Run individual tests:

  pytest <path-to-test-files>
  

Usage

• Example files in /examples/

Examples

Symmetric P1 step pattern

Fig 1: Compared time series and warped sequence.

Fig 2: Cost matrix and optimal warping path.

Fig 3: Accumulated cost matrix and optimal warping path.

Symmetric P0 step pattern

Fig 1: Compared time series and warped sequence.

Fig 2: Cost matrix and optimal warping path.

Fig 3: Accumulated cost matrix and optimal warping path.

References

• Müller, Meinard. Information retrieval for music and motion. Vol. 2. Heidelberg: Springer, 2007. https://doi.org/10.1007/978-3-540-74048-3