The effect of clinical intervention on the speech of individuals with PTSD: features and recognition performances
Abstract
Post-traumatic stress disorder (PTSD) is an anxiety disorder that can occur as a response to traumatic experiences, such as catastrophic events, and can have a detrimental influence on mental wellbeing. Furthermore, PTSD is present in 5-10 % of the population, making it a prevalent disorder in our time, thus necessitating a timely diagnosis and proper treatment. In this paper, we present results for PTSD detection based on speech recordings on a newly collected dataset consisting of 15 participants, including speakers with PTSD and a control group. Moreover, the dataset includes speech data immediately before and after a clinical intervention (i.e., acupuncture-supported psychotherapy), allowing us to examine the effect of a single treatment session. In our experiments, we achieve a best area under the curve (AUC) of 82 % using solely pre-treatment data. Finally, we analyse prominent acoustic patterns of individuals with PTSD compared to the control group.