TOXGB: Teamwork Optimization Based XGBoost model for early identification of post-traumatic stress disorder

Abstract

Post-traumatic stress (PTSD) is considered a clinical issue that influences numerous people from diverse trades all over the world. Numerous research scholars recorded diverse complexities to estimate the severity of the PTSD symptoms in the patients. But diagnosing PTSD and obtaining accurate diagnosing techniques becomes a more complicated task. Therefore, this paper develops a speech based post-traumatic stress disorder monitoring method and the significant objective of the proposed method is to determine if the patients are affected by PTSD. The proposed approach utilizes three different steps: pre-processing or pre-emphasis, feature extraction as well as classification to evaluate the patients affected by PTSD or not. The input speech signal is initially provided to the pre-processing phase where the speech gets segmented into frames. The speech frame is then extracted and classified using XGBoost based Teamwork optimization (XGB-TWO) algorithm. In addition to this, we utilized two different types of datasets namely TIMIT and FEMH to evaluate and classify the PSTD from the speech signals. Furthermore, based on the evaluation of the proposed model to diagnose PTSD patients, various evaluation metrics namely accuracy, specificity, sensitivity, and recall are evaluated. Finally, the experimental investigation and comparative analysis are carried out and the evaluation results demonstrated that the accuracy rate achieved for the proposed technique is 98.25%.

A Novel Stochastic Transformer-based Approach for Post-Traumatic Stress Disorder Detection using Audio Recording of Clinical Interviews

“…The TL approach was used from a speech recognition database, TIMIT, for PTSD identification. They demonstrated how the DBN performed better with the TL technique than the other common methods with an accuracy from 61.83% to 74.99% In [8], the authors suggested a method for PTSD detection that involves three steps: feature extraction, pre-processing or pre-emphasis, and classification. The pre-processing stage receives the incoming audio speech signal first, where the speech is divided into frames.…”