Year: 2026
Type: dissertation
Abstract: An increased demand has created a need for emotional datasets containing both reliable feature extraction from raw physiological signals and consistent machine-readable document that makes datasets discoverable and interpretable. Two tools have been created to address this need. The first is EMGFlow, an open-source Python toolbox that streamlines preprocessing and feature extraction for surface electromyography (sEMG) with an interactive dashboard. It provides a flexible, batch-oriented pipeline that extracts a library of 33 time- and frequency-domain features. The second tool is Datasheets for Emotional Datasets (D4ED), a domain-specific datasheet that encodes metadata often underrepresented or absent in general-purpose dataset documentation. D4ED is designed as a machine-ingestible representation to support dataset indexing, search, and governance in ML/AI repositories. This enables both humans and automated agents to locate, compare, and manage emotion datasets more effectively. Pairing EMGFlows feature extraction with D4EDs structured dataset specification supports researchers in the processing and discovery of datasets.
Author: Conley, D. William
Language: English
Topic: Emotion and Mood Recognition
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7213991758
