Vocal Biomarkers for Cognitive Performance Estimation in a Working Memory Task

The ability to non-invasively estimate cognitive fatigue and workload as contributing factors to cognitive performance has value for planning and decision making surrounding human participation in cognitively demanding situations and environments. Growing evidence supports the use of speech as an effective modality for assessing cognitive fatigue and workload, while also being operationally appropriate in a wide variety of environments. To assess ability to discriminate changes in cognitive fatigue and load from speech, features that measure speech onset time, speaking rate, voice quality and vocal tract coordination from the delta-mel-cepstrum are evaluated on two independent data sets that employ the same auditory working memory task. Feature effect sizes due to fatigue were generally larger than those due to load. Speech onset time, speaking rate and vocal tract coordination features show strong potential for speech-based fatigue estimation.