Research On EEG Signal Acquisition And Classification Algorithm

In recent years,the problem of aging population in our country has become increasingly severe.As the population rises,the incidence of diseases such as cerebral stroke that will cause the patient to lose part of his body control function also increases.The resulting physical motor disfunction caused great inconvenience and mental stress to the patient,but also brought a heavy burden on families and society.How to help patients with effective rehabilitation and regain certain independent living ability is an urgent problem to be solved in today’s society,and it is also a focus of cross-disciplinary research such as rehabilitation engineering and artificial intelligence.EEG is a signal generated by the electrophysiological activity of brain cells and contains abundant brain activity information.By analyzing and evaluating EEG signals quickly and efficiently,we can interpret the brain signals as corresponding commands and build a brain-computer interface system that does not rely on peripheral nerves and muscle tissue to complete communication with the outside world.And this is of great significance for rehabilitation engineering.Focusing on the needs of different application scenarios of brain-computer interface,this paper selects two typical types of EEG signals and studies the signal processing methods of brain-computer interface in the online and offline situations.The following results are obtained:1)The common brain-computer interface systems perform badly in resisting the noise and seem complicated for operation.In this paper,we design a brain-computer interaction system that equipped with Emotiv EPOC+ to overcome such problems.In particular,by combining power spectral density analysis and canonical correlation analysis with different weights,we propose a novel algorithm to improve the recognition rate as high as to 98.6%,and it has high anti-noise ability and extensibility.2)In addition,the brain-computer interface needs to complete a large number of feature engineering and low accuracy when dealing with multi-classification tasks.The deep learning method is used to autonomously extract features of the motion imagery EEG signals,and data enhancement is performed on small sample data sets.Based on the convolutional neural network,a set of high-efficiency and high-precision classifiers is built.Experiments show that the classification method has strong stability and generalization ability.