Optimal Feature Selection using Plant Grow Optimization for Motor Imagery EEG Classification Based on Cascaded Neural Network
The complex state of electroencephalogram signal features deflects the predication rate of the brain-computer interface. The extraction of features from the EEG signal using stationary wavelet transforms function. The stationary wavelet transform function decomposed the layers of sub-bands of EGG signals. The extracted features are a combination of real signal and noise signal and the noise-signal degraded the performance of the EEG classification. For the removal of noise signal form extracted features using plant grow optimization algorithm, the plant grows optimization algorithm filter the noise feature from extracted data. The nature of the plant grows optimization algorithm is impulsively related to the nature of the EEG signal. Various authors proposed an EEG classification algorithm for the prediction of human diseases related to the control system of the brain. The classification algorithm faced a challenge of optimal feature selection and mapping of assign class for the predication. In this paper proposed cascaded neural network models for the classification of EEG signals. The cascaded neural network model is a combination of a Self-organized map network and a radial biased function. The RBF neural network is a supervised network model, and the SOM model is unsupervised. The SOM model performs the task of a grouping of signals, and RBF neural network model works for the predication of pattern for the classification process. The proposed algorithm simulates in EGG LAB with MATLAB 2016Ra. For the validation used sampled data of EEG signals from the reputed data source.