Electroencephalogram multi-domain feature extraction method based on multivariate variational mode decomposition
A technique of variational modal decomposition and feature extraction, which is applied in the field of Ming Dynasty and can solve the problems of ignoring time domain features and spatial distribution components.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2020-09-04
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention belongs to the field of pattern recognition, and is aimed at motor imagery electroencephalogram (EEG), and proposes to use multivariate variational mode decomposition (MVMD) to carry out adaptive decomposition on the original EEG multi-channel data, and then obtain the intrinsic mode function from the decomposition The (IMF) component extracts the time-domain features and nonlinear dynamic features of the signal. At the same time, the IMF components are combined to construct a new signal matrix, and the common space pattern (CSP) method is used to extract the spatial features of the reconstructed signal to perform time-domain, Combination of nonlinear dynamics and spatial features features and classification methods. Background technique
[0002] Brain-computer interface technology (BCI) is a human-computer interaction system that directly communicates information between the human brain and the outside world without relying on the normal...
Examples
Embodiment Construction
[0022] The MVMD-based EEG multi-domain feature extraction of the present invention will be described in detail below in conjunction with the accompanying drawings, as figure 1 , the implementation of the present invention mainly comprises 6 steps: (1) collect multi-channel EEG signal and preprocessing, (2) carry out MVMD decomposition to multi-channel EEG signal, obtain some intrinsic mode function IMF components, (3) According to the Hilbert spectrum analysis of each IMF component, the instantaneous energy mean value feature of the EEG signal is obtained, (4) extract the multi-scale sample entropy feature for each IMF component, (5) extract the variance vector from the new signal matrix composed of components through the CSP algorithm , (6) Combining the features obtained in (3)(4)(5) into the classifier for classification to obtain the result.
[0023] Each step will be described in detail below one by one.
[0024] Step (1): In this example, the BCI Competition II Dataset ...