KACZMAR SPATIO Temporal Nelder Mead Multilayer Perceptrons for Stress Detection Using EEG Signals
Muhammadu Sathik Raja M S, Arun Raaza, Mee na, Farida Virani · International Journal of Engineering Trends and Technology · 2024
Stress is an emotion that people encounter when they are extremely loaded and encounter trials and tribulations while carrying out day-to-day chores. Stress influences individual health seriously, like soaring blood pressure, heart disease, cardiovascular disease, and even lead to stroke. As a result, early stress detection becomes helpful to keep an eye on health-related issues caused by stress. Electro Encephalography (EEG) signal based system assists in identifying the different disorders and disabilities. Hence, there is a requirement for early stress detection using EEG signals that are accurate, precise, and reliable. This is resolved in the proposed method by introducing Kaczmar Spatio Temporal Nelder Mead Multilayer Perceptrons (KST-NMMP) that can accurately classify and detect the stress level. In this KST-NMMP method, deep learning using multilayer perceptrons is employed for early stress detection. It is split into four layers, i.e., one input layer, two hidden layers, and one output layer. The input EEG signals obtained from the subjects are provided in the input layer. Next, in the first hidden layer, the artifacts present in the raw EEG signals are filtered out; thus, the stress detection time can be reduced. After noise reduction, the spatial and temporal domain features are extracted from EEG signals; thus, stress detection overhead can be reduced significantly. Finally, stress level classification and detection at an early stage are performed in the second hidden layer employing spatial and temporal features using the Nelder Mead activation function. This proposed KST-NMMP method ensures accurate classification outcome which leads to improvement both in terms of precision and recall significantly. The overall implementation is performed in the Matlab programming language. Finally, the performance is evaluated and compared with the conventional method in terms of precision, recall, stress detection time, and stress detection overhead.