Inferring Autonomic Nervous System Activation from Noisy Wearable Electrodermal Activity Data with the Goal of Investigating the Relationship Between Vocal Hyperfunction and Emotional Arousal

Date

2020-09-29

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Vocal hyperfunction is a condition characterized by chronically excessive or unbalanced vocal muscle recruitment. Patients with non-phonotraumatic vocal hyperfunction experience higher than normal levels of psychological stress when speaking, which could imply a relationship between high arousal and vocal hyperfunction. As such, investigating the relationship between arousal level and voice activity may provide insights into the prevention, diagnosis and treatment of vocal hyperfunction. Arousal information can be inferred from eletrodermal activity data. As a measure of electrodermal activity, skin conductance reflects the stimulation from the autonomic nervous system on eccrine sweat glands due to arousal. However, extracting neural stimuli from skin conductance measurements is challenging, as the underlying physiological system is unknown. Moreover, artifacts originating in real-world settings can corrupt the skin conductance signal, making portions of the signal unsuitable for analysis. We investigate two published automatic methods for identifying electrodermal activity segments suitable for analysis. We also visually inspect the data and compare with automatic methods. Of the current methods available for identifying suitable regions of electrodermal activity data for analysis, visually selecting regions is the most reliable and the most conservative. Then, we isolate the satisfactory segments of electrodermal activity data for analysis. Using a generalized-cross-validation-based block coordinate descent approach for sparse deconvolution, we recover underlying neural stimuli and model parameters from the skin conductance signal. In future work, we plan to investigate the relationship between voice data and the underlying neural stimuli and model parameters recovered from electrodermal activity data. This project was completed with contributions from Andrew J. Ortiz and Daryush D. Mehta from Massachusetts General Hospital.

Description

Keywords

Citation