Wearable ECG sensor based cognitive stress detection during physical activity using Multifractal Detrended Fluctuation Analysis

Real-time cognitive stress detection using wearable sensors is challenging due to the overlapping nature of physiological responses with physical activity induced artifacts. This necessitates the need for advanced signal processing methodologies capable of identifying cognitive stress from movement-induced and breathing artifacts. In this work, multifractal features are extracted and analyzed to identify its potential in differentiating cognitive stress during physical activity. The wearable ECG signals are obtained for two conditions: cognitive stress at rest and cognitive stress during physical activity, which is pre-processed to remove powerline interference, high-frequency noises, and baseline wandering. The multifractal characteristics are analyzed by extracting the fluctuation function, generalized Hurst exponent and mass exponent. From the fluctuation function, it is observed that rapid variations at smaller scales in cognitive stress are attenuated during physical activity. The generalized Hurst exponent shows a decreasing trend with increasing order of q, exhibiting multifractality behavior in both the considered conditions. Similarly, mass exponent also exhibits a non-linear concave curve conforming to the multifractal property of the signal. Further, multifractal features are extracted and subjected to statistical t-test to identify the significance of these features. The extracted features, such as spectral width, area under spectrum, and broadness, show signification variation compared to cognitive stress during physical activity. This indicates that cognitive stress has higher singularity strength and different scaling behavior across time scales. Additionally, the extracted features are also observed to be statistically significant with p<0.05 in differentiating the two conditions. These findings support the effectiveness of the proposed multifractal-based approach for distinguishing cognitive stress under physical activity and could be integrated into existing wearable sensor platforms for real-time, artifact-resilient stress detection.