UC Berkeley PhonLab Annual Report

This is a short paper comparing two approaches to head correction for Electro-MagneticArticulography (EMA) data collected with the Northern Digital Instruments “Wave” system. Inboth of these approaches, it is necessary to translate and rotate the sensor locations to theocclusal coordinate system. We found that point tracking error is greater by as much as doublewith the built-in NDI head correction method, compared to a three-sensor head correctionalgorithm. However, we conclude that the data are comparable, and that the two-sensor NDImethod is acceptable for phonetic research. A Python library for head correction was developedfor this work, and is available on github.com.