Application for audio data modelling and machine learning on biomedical audio data.
[]
soundofnavi has a dual purpose. It serves as a data modelling package for biomedical (lung) audio data, in synergy with a machine learning pipeline emphasizing scalability and interpretability.
The library structures any lung sound dataset into a neat object oriented design, where concepts of Datasets, Patients, Recordings and Slices are tied together for audio loading, preparation, augmentation, and advanced analysis.
The fun comes in with a machine learning framework applicable on multiple tasks (i.e., pneumonia, crackles/wheezes) with extensive metrics, visualizations, etc, not only on a piece of audio at fixed time, but across any desired critera (i.e., by dataset; by patients of a certain demographic) over time.
Available on PyPI at {TODO} and GitHub at https://github.com/arachid1/soundofnavi
Official documentation at https://soundofnavi.readthedocs.io/en/latest/