Electroencephalogram signal feature recognition method under motor imagery state

Abstract

The invention discloses an electroencephalogram signal feature recognition method under a motor imagery state. The electroencephalogram signal feature recognition method includes the following steps:S1, collecting electroencephalogram signal data information of a subject under the motor imagery condition; S2, calculating energy spectrums of electrodes of electroencephalogram signals by using a Welch method; S3, setting personalized information of the optimal electrode of the electroencephalogram signals, and selecting the electrode with the highest classification accuracy for different subjects; S4, extracting feature values influencing imagery of the electroencephalogram signals by using a synchronizing / desynchronizing method; and S5, performing feature classification on the extracted electroencephalogram signals by using the optimal classification function, and optimizing the classification process by using a three-stage type classification method based on a support vector machine.The electroencephalogram signal feature recognition method under a motor imagery state classifies the electroencephalogram signals by calculating the energy spectrum and the feature value informationof the electrodes of the electroencephalogram signals, wherein electroencephalogram signal feature extraction and classification techniques through motor imagery can be used in the field of neurological rehabilitation.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to a method for identifying features of electroencephalogram signals in a motor imagery state. Background technique [0002] Motor imagery EEG feature extraction and classification technology can be used in the field of neurorehabilitation, for those who are paralyzed or severely motor-impaired, especially those with complete brain functions but unable to move (such as multiple sclerosis, amyotrophic lateral cord patients with diseases such as sclerosis), provide a new way to communicate with the outside world, bypass damaged neurons, and regain control of a limb or prosthesis. This technology can also be used in fields such as transportation, military affairs, leisure and entertainment, etc., and has wide application prospects. At present, the main feature extraction methods of EEG signals are time-domain methods, multi-dimensional statistical analysis methods, frequency...

AI Technical Summary

Problems solved by technology

[0005] Inducing EEG signals requires additional stimulation devices and depends on a certain sensory pathway of the user, so the scope of application is limited

In addition, it is not suitable for long-term continuous use, otherwise it will cause the subject to adapt to related events or physical fatigue, which will cause changes in related potentials. This contradiction has yet to be resolved.

The EEG signal of the spontaneous system is completely derived from the subject's spontaneous EEG, and does not require external stimulation, so it does not require the subject's sensory nerve function, but due to its non-stationarity, it is easily affected by the environment and emotions, and the characteristic representation Not significant, so this system has high requirements for signal processing methods, and the current recognition accuracy is low

[0006] 2. The signal collection method needs to be improved

[0007] The EEG signal is a complex non-stationary signal, very weak, and easily disturbed by the outside world. Therefore, how to design a more reasonable experimental plan and improve the signal-to-noise ratio needs to be solved.

At present, the collection of EEG signals is mainly through implantable electrodes and external electrodes. The external electrodes can be pasted one by one or wearing electrode caps. The number of electrodes used is usually a lot, and the collection of EEG data cannot be done by the recipient. The tester does it alone

In addition, the optimal electrode position for each subject is different, and the individualization and precision of the electrode position cannot be realized at present, which affects the signal test accuracy

[0008] 3. EEG signal pattern recognition methods need to be improved

[0013] When leaving the laboratory environment, subject to the interference of the external environment and its own conditions, the spontaneous EEG signals generated by the subjects will continue to change and be easily affected by artifacts. The system not only needs to accurately judge task execution and idle status, but also To quickly and accurately extract signal features, the current brain-computer interface technology cannot meet the requirements, and subjects cannot flexibly control peripheral devices through the brain-computer interface in a real environment

Images