Feature Extraction of Motor Imagination EEG Signal Based on VHTSM

Strong randomness of brain electric signal, low am-plitude, the characteristics of non-stationary characteristics are extracted, because of strong noise interference, makes less useful feature vectors, which affects the subsequent research, therefore, this paper proposes a detection algorithm based on variational Hilbert transform spectrum matrix, first of all, the original signal is split into multiple child signals, and the various sub signal variation mode decomposition, Energy concentrated into consideration, based on the frame for the unit, construct the brain electrical signal spectrum matrix, and put forward a method of 3d visualization analysis effectively find the reproduction matrix of the difference between signal and noise, then use power spectral density of signal feature extraction and support vector machine (SVM) classification, finally the experimental results show that the average accuracy of 81.22%.