A Dynamic Artificial Neural Network for EEG Patterns Recognition

EEG—Electroencephalogram is the register of the brain activity, usually obtained from a non-invasive method that uses metal electrodes connected on the subject scalp, according to International 10–20 System of Electrode Placement. The analysis of an EEG signal can reveal a patient’s state of healthiness related to some neurological disorders, such as Alzheimer, Parkinson and Epilepsy. In this paper an artificial neural network with dynamic sampling strategy of an EEG signal based on backpropagation algorithm for recognition of brain waves in Alpha, Beta, Theta and Delta patterns was implemented with accuracy of 95%. As these waves are related to the usage of brain regions involved according to the instant task and its specificity, such as cognitive tasks, spelling and writing tasks, sport practices tasks or even resting, the result of its recognition followed of qualitative analysis can be used for applications in diagnosis of neurological diseases.