A Novel Spatial-Spectral Signal Processing Method for Rehabilitation EEG Data Analysis of Stroke Patients

The Spatial Spectral signal processing method is a method for analyzing Motor Imagery (MI) Electroencephalography (EEG) of stroke patients. EEG analysis is used to pre-define process of the channels configuration and frequency band classifying the EEG for stroke patients. EEG analysis based data was recorded by a 16 channel. The EEG is performed heavily depends on the selection of a time interval. That pattern may have gradually changed during rehabilitation. The main issues of the EEG analysis report of stroke patients is formation of some unknown channels which is contaminated with more noisy and non stationery signals. In this paper, classical CSP based method is improved for data analysis during motor recovery motor imagery EEG patterns of stroke patients which changes that the rehabilitation training process. This is methods is based on variable preconditions and introduced a new heuristic supervisor of stochastic gradient boost strategy for training weak classifiers of the spatial spectral during rehabilitation method. Real-world datasets were collected from famous BCI competitions for training and test dataset using this method. This method has been implemented for the channel and frequency using a sliding window process and then trained the weak learners for that boosting signals. Three different datasets are tested and recorded for including the healthy people and stroke patients. The test accuracies are obtained for each subject in Brain Computer Interface (BCI) competition of the competing tests are PSD, SR, CSP, RCSP, SBCSP, CSSP, and CSSSP. This new method has been evaluated for its performance benchmark based on the session to-session transfer rate. The experimental results using various simulations show that the proposed algorithm is outperformed with conventional systems as well as reducing the computational complexity.