EEG based Mental Stress Detection using Deep Learning Techniques

The importance of identifying stress in living in a fast culture cannot be overstated. The electrical bio-signals produced by the brain are read out using an electroencephalography (EEG) method. Yet, owing to their intricacy, EEG signals can only be deciphered by a physician with extensive training in this area. In order to identify human stress, this research offers a DWT-based hybrid deep learning model that utilizes convolution neural networks and Bidirectional Long Short-Term Memory (CNN-BLSTM). Helpful for psychiatrists, psychologists, and other medical professionals who need to assess a patient’s stress levels. Mental math stress is detected with the use of the Physionet EEG dataset. Using Discrete Wavelet Transform, noise has been eliminated and split into four levels from multi-channel (19 channels) EEG data (DWT). After decomposition, the signals are subjected to an automated feature selection technique using a CNN. At last, stress levels are categorized using BLSTM. The suggested model’s accuracy is evaluated in comparison to that of CNN-based LSTM and prior work. Based on the data, it’s clear that the suggested mixed model is superior to others in terms of classification accuracy (99.20 percent was attained). Moreover, the suggested model is validated with a classification results of 98.10% using the stratification tenfold cross-validation method.