Classification of speech under stress and cognitive load in USAR operations

Marcela Charfuelan Oliva, Geert-Jan Kruijff

In: Proceedings of ICASSP. International Conference on Acoustics, Speech and Signal Processing (ICASSP-2013) May 26-31 Vancouver BC Canada IEEEXplore 2013.


This paper presents the classification of speech under stress and cognitive load in speech recordings of Urban Search and Rescue (USAR) training operations. The type of stress encountered in the USAR domain, more specifically in the human team communication, includes both physical or psychological stress and cognitive load. We were able to annotate and identify these two types of stress in recordings of real USAR raining operations. Different acoustic features are extracted at full and subband level, SVM and adaptive GMMs are used as classifiers. Two strategies to improve the classification of speech under stress, in particular physical stress, are proposed. We have achieved a classification accuracy of 74% for three very unbalanced classes (physical stress, cognitive load and neutral), with 82% classification of physical stress.


German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz