EMGFlow

The open workflow for EMG signal processing and feature extraction.

EMGFlow is a Python package for researchers and clinicians to engage in signal processing. EMGFlow provides a broad range of functions to meet your EMG signal processing needs, without prescribing a specific workflow. EMGFlow follows open standards of data processing, such as CSV files and Pandas data frames to allow easy integration. With functions to extract 32 different features according to your needs, EMGFlow provides a uniquely deep feature extraction.

EMGFlow also includes an easy method for producing detailed graphs of EMG signals in large quantities.

The simple navigation menu allows you to switch between which files you see as well as which layers are overlayed:

Statement Of Need

While several packages exist for processing physiological and neurological signals, support for surface electromyography (sEMG) has remained limited. Many packages lack a comprehensive set of features that can be extracted from sEMG data, leaving researchers to use a patchwork of tools. Other packages are orientated around event detection in individual recordings and use GUI-based workflow that requires greater manual intervention. While this design works well for single person recordings, it complicates the extraction of features from large datasets, such as those common to machine learning.

The EMGFlow package a portmanteau of EMG and Workflow - fills this gap. Our package provides a flexible pipeline for the extraction of a wide range of sEMG features, with a scalable design that is suited to the processing of large datasets.

Example

As a quick example, the following will create a feature file, starting with a folder of raw data: ```python import EMGFlow

Get path dictionary

pathnames = EMGFlow.makepaths()

Load sample data

EMGFlow.makesampledata(path_names)

Sampling rate

sampling_rate = 2000

Filter parameters

notchvals = [(50, 5)] bandlow = 20 bandhigh = 140 smoothwindow = 50

Columns containing data for preprocessing

cols = ['EMGzyg', 'EMGcor']

1. Apply notch filters

EMGFlow.notchfiltersignals(pathnames['Raw'], pathnames['Notch'], samplingrate, notchvals, cols)

2. Apply bandpass filter

EMGFlow.bandpassfiltersignals(pathnames['Notch'], pathnames['Bandpass'], samplingrate, bandlow, band_high, cols)

3. Apply smoothing filter

EMGFlow.smoothfiltersignals(pathnames['Bandpass'], pathnames['Smooth'], smooth_window, cols)

4. Extract features

df = EMGFlow.extractfeatures(pathnames, sampling_rate, cols) ```

Documentation

General: - EMG processing background - EMGFlow processing pipeline overview - FileAccess module API - PreprocessSignals module API - ExtractFeatures module API - OutlierFinder module API - PlotSignals module API

Examples: - Processing pipeline examples

Installation

EMGFlow can be installed from PyPI: python pip install EMGFlow

Once installed, the package can be loaded as follows: python import EMGFlow

Project dependencies can be seen here.

Contributions

Contributions and community guidelines can be seen here.

Citations

This package can be cited as follows:

bibtex @software{Conley_EMGFlow_2024, author = {Conley, William and Livingstone, Steven R}, month = {03}, title = {{EMGFlow Package}}, url = {https://github.com/WiIIson/EMGFlow-Python-Package}, version = {1.0.17}, year = {2024}, note = "{\tt william@cconley.ca}" }

If you are using a different version of EMGFlow, change the version tag to the version you are using.