https://github.com/hahnec/multimodal_emg

Multimodal Exponentially Modified Gaussians with Optional Oscillation
https://github.com/hahnec/multimodal_emg

acoustic-features acoustics exponentially-modified-gaussian feature-engineering gaussian-mixture-models mixture-model multimodal oscillation oscillations regression regression-models superposition

Last synced: 1 day ago
JSON representation

Multimodal Exponentially Modified Gaussians with Optional Oscillation

• Host: GitHub
• URL: https://github.com/hahnec/multimodal_emg
• Owner: hahnec
• Created: 2022-07-24T15:50:20.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2023-03-29T11:29:46.000Z (over 1 year ago)
• Last Synced: 2024-01-27T15:42:46.045Z (10 months ago)
• Topics: acoustic-features, acoustics, exponentially-modified-gaussian, feature-engineering, gaussian-mixture-models, mixture-model, multimodal, oscillation, oscillations, regression, regression-models, superposition
• Language: Jupyter Notebook
• Homepage:
• Size: 2.4 MB
• Stars: 5
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Multimodal Exponentially Modified Gaussians

[![arXiv paper link](https://img.shields.io/badge/paper-arXiv:2209.12202-red)](https://arxiv.org/pdf/2209.12202.pdf)

## Quick Facts

- multiple asymmetric Gaussian distributions for the univariate case

- optional oscillation term for wave approximations

- based on [analytical derivation](./docs/exp_mod_gauss_wave_partial_derivative.pdf)

- accepts ```numpy``` as well as ```torch``` data types

## Kick Start

Below is a code excerpt for fitting multi-modal skewed Gaussian distributions:

```python

from multimodal_emg import gaussian_envelope_model, emg_envelope_model, emg_wave_model

from multimodal_emg.regression.derivatives import gaussian_jac, emg_jac, oemg_jac

# multimodal optimization

p_star, result = multimodal_fit(

    data,

    features = [[1, 24, 2, 0],[.5, 48, 3, -1]], # amplitude, location, spread, skew

    components = 2,

    x = x,

    fun = emg_envelope_model,

    jac_fun = emg_jac,

)

print(p_star)

import matplotlib.pyplot as plt

plt.plot(result)

plt.show()

```

## Oscillating Regression

The oscillation regression can be found in the accompanied [Jupyter Notebook](./single_synth_memg_example.ipynb) which yields the below result:

[![Oscillating Multi-Modal EMG](./docs/figs/denoising_emg.svg)](./single_synth_memg_example.ipynb)

## Citation

```

@inproceedings{Hahne:2022,

    author = {Christopher Hahne},

    title = {Multimodal Exponentially Modified Gaussian Oscillators},

    booktitle= {2022 IEEE International Ultrasonics Symposium (IUS)},

    address={},

    month={Okt},

    year={2022},

    pages={1-4},

}

```

## Acknowledgment

This research is funded by the Hasler foundation under project number 22027.