A method and system for multi-classification of brain consciousness based on support vector machine

Abstract

The invention discloses a brain consciousness multi-classification method and system based on a support vector machine, wherein the method includes: collecting the brain blood oxygen original data sequence of the subject's brain motor area during the brain consciousness movement period; preprocessing the sequence And extract feature value to construct feature vector, as training sample; Obtain training sample respectively for all k kinds of consciousness movement categories; According to the quantity k of consciousness movement category, construct k(k-1) / 2 SVM classification models; For each The SVM classification model obtains corresponding training samples from the training set to form a corresponding sub-sample set; adopts the sub-sample set training to obtain the SVM binary classifier of the corresponding combination category, and combines to obtain the multi-classifier of the consciousness movement category; obtain the waiting list according to the above method The feature vector of the subject is input to the multi-classifier of the consciousness movement category to identify the consciousness movement of the subject. The invention can realize online real-time multi-category recognition for unknown consciousness movement categories.

Description

technical field [0001] The invention relates to the technical field of brain-computer interface, in particular to a method and system for multi-classification of brain consciousness based on support vector machines. Background technique [0002] The existing non-invasive BCIs include electroencephalography (EEG), magnetoencephalography (MEG), functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS). When using MEG and fMRI technology, the subjects are required not to have any movement, so it is not suitable for application in scenes with a lot of movement; compared with MEG and fMRI, EEG has a higher time resolution and is the best choice for studying brain-computer interfaces. It is an important method, but before using it, the subjects need to be trained to understand the operation process and purpose, and the operation process is long and easy to fatigue the subjects, and the signal-to-noise ratio collected in the exercise experiment ...

AI Technical Summary

Problems solved by technology

[0006] Based on the technical problems of low precision of existing multi-classification of brain consciousness movements and long processing time for offline recognition, the present invention provides a method and system for multi-classification of brain consciousness based on support vector machines, which realizes online classification of various brain consciousness movements real-time classification

Images