This is the repository sharing the code for the work done on the paper "Surfboard: Audio Feature Extraction for Modern Machine Learning" at INTERSPEECH 2020. The paper can be found on arXiv. The Surfboard repository can be found here.
Contents
Comparison.ipynb: Comparison of common features (local jitter, DDP jitter and local shimmer) between the Surfboard, OpenSMILE and Praat frameworks.ParkinsonsClassification.ipynb: Example classification task using Surfboard features, highlighting the importance of keeping train/test sets disjoint by participant and matching age/gender to create meaninful validation metrics.ReferenceValues.ipynb: Calculation of Surfboard reference means, medians, standard deviations and median absolute deviations on a 40-hour subset of LibriSpeech.