A New Scheme of EEG Identification Based on MFDC

To solve the problem of the existing Electroencephalogram (EEG) signal identification scheme including incomplete de-noising, over-complexity and weak robustness because of not taking into account the particularity of EEG data, a new identification algorithm named MFDC (MFCC, DWT, CNN) is proposed based on the connection between audio signal and EEG signal. In this scheme, three method MFCC, DWT, CNN are merged and improved to process the EEG signal. Firstly, the original EEG signal is preprocessed by Discrete Wavelet (DWT) for objectives include discrete wavelet de-noising, pre-emphasis, framing and key-frame screening. Then, after the MFCC of EEG signals is extracted and normalized, the data processing model based on Convolutional Neural Network (CNN) is used to process the EEG feature. Finally, the identification should be detected by similarity determination. In order to verify the feasibility of the key-frame filtering in the preprocessing, the algorithm is analyzed from the aspects of validity and characterization. At the same time, in order to verify the feasibility of the scheme, it is analyzed from the perspectives of robustness and performance. The experiment results show that the key-frames selecting has strong representativeness, and the identification scheme has the advantages of low complexity, high accuracy and strong robustness.