Mental Workload Assessment from ECG and PPG Signal Using Machine Learning Techniques

Chronic stress is closely associated with the onset of mental health disorders and cardiovascular conditions, underscoring the need for effective stress monitoring to support overall well-being. Despite this, many individuals frequently ignore stress indicators, which can lead to significant health complications. Recent neuroscience research identifies mental stress as a major health crisis of the 21st century, yet standardized methods for its assessment are lacking. Although cortisol levels in saliva have traditionally served as an indicator, advancements in wearable technologies have drawn attention to physiological markers, such as heart rate variability (HRV), for stress evaluation. This study proposes a supervised machine learning approach to detect mental stress by analyzing HRV. Data were gathered from 22 participants who completed the N-back task, a cognitive workload test, while their electrocardiogram (ECG) and photoplethysmography (PPG) signals were recorded. HRV features were extracted and labeled using NASA-TLX scores to quantify mental workload. Various machine learning models with cross-validation were applied, yielding high accuracy rates of 96% and 88% with the random forest classifier, demonstrating a promising approach for effective stress detection.