Abstract
<jats:p><p>This article presents the development and validation of a method for identifying acoustic predictors of stress in operators of complex technical systems, based on Kohonen Self-Organizing Maps (SOM). The relevance of this research is driven by the need to enhance the operational safety of complex technical systems &mdash; particularly aircraft &mdash; through objective, non-contact monitoring of personnel's psychoemotional state. The study was conducted as part of a research project aimed at designing an advanced civil aircraft cockpit equipped with an intelligent crew support system. A key requirement for this cockpit is the provision of continuous, non-invasive monitoring of pilots' functional state to enable timely detection of critical changes associated with stress and fatigue. The proposed approach involves extracting a nine-dimensional acoustic feature vector from the operator's speech signal &mdash; comprising fundamental frequency (F0), spectral centroid, spectral bandwidth, spectral roll-off, and five peak frequencies of the power spectrum &mdash; followed by classification of the current state as either "normal" or "stress" using a trained Kohonen map. To verify the proposed approach, a software model called "AudioStressPredictor" was developed, implementing the full processing pipeline: from creating a personalized operator profile and training the model to real-time stress level monitoring. Experimental testing was carried out, and the results confirm the viability of the proposed approach and its potential for application in aviation training centers, as well as as one of the key components of onboard decision support systems for advanced aircraft.</p></jats:p>