Enhanced Driver Stress Prediction from Multiple Biosignals via CNN Encoder-Decoder Model

In this work, we present PhysioFuseNet, a novel framework designed to enhance driver stress state classification. PhysioFuseNet integrates a CNN-based encoder-decoder model with multimodal biosignal fusion. Using a driving simulator, different multimodal signals were acquired, namely electrocardiography, electrodermal activity, photoplethysmography, and respiration rate from (N = 25) healthy subjects. The experiment is of 35 minutes duration and contains different stress states (baseline (5 minutes), while normal, cognitive, and emotional sessions for 10 minutes). Multimodal features are extracted and employed in an encoder-decoder network. Extracted encoder features are combined through intermediate fusion and fed to support vector machine (SVM) and random forest (RF) classifiers. Experimental results demonstrate the efficacy of our approach, outperforming previous methods by achieving accuracies of 0.95 and 0.94 for SVM and RF, respectively. Notably, the framework excels in classifying emotional and cognitive stress states. In summary, the proposed framework could be useful in stress assessment in real-time and clinical conditions.