Electroencephalogram signal feature extraction method based on self-attention mechanism

Abstract

The embodiment of the invention discloses an electroencephalogram signal feature extraction method based on a self-attention mechanism. In a specific embodiment, the method comprises the following steps: obtaining an original electroencephalogram signal of a subject through a multi-channel electroencephalogram acquisition device, and preprocessing the original electroencephalogram signal; creating a filter based on a self-attention mechanism, and extracting a first feature vector of the to-be-classified motor imagery matrix by using the filter; obtaining a second feature vector of the to-be-classified motor imagery matrix based on a Riemannian manifold feature extraction algorithm; and obtaining a feature vector of the to-be-classified motor imagery matrix based on the first feature vector and the second feature vector.

Description

technical field [0001] The invention relates to the field of human brain recognition. More specifically, it relates to an EEG feature extraction method based on a self-attention mechanism. Background technique [0002] As a new type of human-computer interaction system, brain-computer interface involves many disciplines and fields. The research on the brain-computer interface system at home and abroad has gone through many years, but so far only a small part of the brain-computer interface system is practical, and most of them are still in the theoretical and laboratory stage. important role in the field. For example, in the medical field, brain-computer interfaces can solve problems in the daily life of disabled people. In the field of aerospace, because astronauts are in a special environment in outer space, it is inconvenient to move and work. With the help of the brain-computer interface system, robots can be used to assist in completing tasks and ensure the safety of...

AI Technical Summary

Problems solved by technology

[0004] The original EEG signal presents non-stationary characteristics and low signal-to-noise ratio in terms of data characteristics, resulting in weak regularity of the EEG signals of the same subject at different time periods, and the EEG signals brought about by individual differences of different subjects are inconsistent. The rules are stronger, how to improve and ensure the correct rate of EEG signal classification has always been a difficult point in the research of brain-computer interface systems

Images