The dynamic collective Kalman filter framework for enhanced EEG-based mental stress detection

Analyzing nonstationary EEG signals for mental stress detection presents significant challenges, requiring advanced frameworks capable of adapting to dynamic noise conditions and signal variability. Conventional filter, such as Adaptive Kalman Filter (AKF) and Locally Adaptive Cooperative Kalman Smoothing (LACKS), techniques may prove inadequate, exhibiting excessive sensitivity to noise variations or computationally intensive. This research offers the Dynamic Collective Kalman Filter (DCKF) Framework, which employs exponential filter to dynamically update noise covariance matrices Q (process noise) and R (measurement noise), providing progressive updates that reduce the influence of tiny fluctuations. DCKF framework ensures that the estimations are consistent, thanks to collective filtering. This will make sure that the averaged outputs of different estimators have eliminated all kinds of switching artifacts. Testing on the SAM:40 EEG dataset, including such tasks as arithmetic, mirror-image, and Stroop tests, provides accuracy of 90 ± 1.86%, precision 92 ± 3.38%, and recall 85 ± 2.75%, and gives testimony to being both effective and efficient. These results show that DCKF is a reliable and scalable method for EEG-based stress detection, combining good accuracy with computational efficiency.