A semi-supervised EEG sleep staging method based on multi-domain features

Abstract

The invention relates to a semi-supervised EEG signal sleep staging method under multi-domain characteristics, belonging to the field of brain-computer interface. In this method, the original EEG signal is processed by combining the ant colony algorithm and the semi-supervised Bayesian classification method to obtain the original feature set under multi-domain features, which is optimized by the ant colony algorithm and an optimal Feature subsets, and use the improved semi-supervised Bayesian classification method, using the optimal feature subset as the classification feature, combined with active learning strategies to classify it in stages. This method can not only effectively realize sleep staging, avoid the waste of computing resources caused by low-efficiency feature attributes and blind search process, but also reduce the number of labeled samples. By comparing with various sleep staging methods proposed in recent years, the results It shows that this method can not only achieve the staging accuracy effect of the mainstream algorithm, but also greatly reduce the demand for labeled samples and avoid the occurrence of manual misjudgment.

Description

technical field [0001] The invention belongs to the technical field of brain-computer interface, and relates to a semi-supervised EEG signal sleep staging method under multi-domain characteristics. Background technique [0002] The sleep mechanism is closely related to the health status of the human body, and plays a key role in many important body activities of the human body. With the rapid development of society, increasing competition and accelerated pace of life, people's life pressure is gradually increasing, and the incidence of insomnia diseases is increasing year by year, which has attracted the attention of medical circles at home and abroad. In the process of treating insomnia, sleep EEG can objectively present the characteristics of certain sleep processes. Through the research and analysis of EEG signals, a large amount of useful information about the human body can be obtained, which is useful for the study of human brain function and disease diagnosis. Theref...

AI Technical Summary

Problems solved by technology

However, due to the instability and non-linear characteristics of EEG signals, the effective features that can perfectly characterize the changes of sleep stages should not be obtained through a single domain feature parameter, and the use of a large number of feature parameters with different properties will also bring Problem: A large number of feature parameters will lead to the expansion of the feature dimension space, reduce the calculation efficiency, and lead to the occurrence of "dimension disaster"

In addition, when the effective feature parameters are obtained, there is also a defect in using a classifier to classify them: traditional sleep stage classification methods often focus on the classification accuracy value, but ignore the number of labeled samples in the classification samples used to train the classifier The pressure brought to the work of sample collection and labeling, because the type classification of labeled samples needs to rely on the manual labeling of experts. If the amount of data is too large, it will not only lead to an increase in the cost of manual labeling, but also lead to the probability of classification errors if the work is long. rise, thus affecting the classification accuracy

Images