Multi-class EEG signal classification with statistical binary pattern synergic network for schizophrenia severity diagnosis

Electroencephalography (EEG) is a widely used medical procedure that helps to identify abnormalities in brain wave patterns and measures the electrical activity of the brain. The EEG signal comprises different features that need to be distinguished based on a specified property to exhibit recognizable measures and functional components that are then used to evaluate the pattern in the EEG signal. Through extraction, feature loss is minimized with the embedded signal information. Additionally, resources are minimized to compute the vast range of data accurately. It is necessary to minimize the information processing cost and implementation complexity to improve the information compression. Currently, different methods are being implemented for feature extraction in the EEG signal. The existing methods are subjected to different detection schemes that effectively stimulate the brain signal with the interface for medical rehabilitation and diagnosis. Schizophrenia is a mental disorder that affects the individual's reality abnormally. This paper proposes a statistical local binary pattern (SLBP) technique for feature extraction in EEG signals. The proposed SLBP model uses statistical features to compute EEG signal characteristics. Using Local Binary Pattern with proposed SLBP model texture based on a labeling signal with an estimation of the neighborhood in signal with binary search operation. The classification is performed for the earlier-prediction shizophrenia stage, either mild or severe. The analysis is performed considering three classes, i.e., normal, mild, and severe. The simulation results show that the proposed SLBP model achieved a classification accuracy of 98%, which is ~12% higher than the state-of-the-art methods.