Method and system for marking data types of EEG signals in awake state

Abstract

The invention relates to an electroencephalogram data type annotation method and system in a conscious state, wherein the method comprises the following steps that after a user starts a sleeping process, an electroencephalogram sample of the user is collected; the electroencephalogram sample is subjected to wavelet decomposition; the signal rebuilding is performed according to the wavelet coefficient of the set low frequency band to obtain an electroencephalogram; the sample entropy of the electroencephalogram is calculated, and is compared with a pre-calculated sample entropy threshold vale; if the sample entropy is greater than the sample entropy threshold value, the signal type of the electroencephalogram sample is annotated into the conscious state. By using the technical scheme, the method and the system have the advantages that the electroencephalogram is prevented from being interfered; the conscious state of the electroencephalogram is accurately detected; in addition, the effective data type annotation is performed, so that the recognition accuracy of a personnel classifier trained by using the annotated electroencephalogram sample is higher; the reliability of the detection result of the personnel sleeping stage in the later period is also improved.

Description

technical field [0001] The invention relates to the technical field of sleep aids, in particular to a method and system for labeling data types of brain electrical signals in an awake state. Background technique [0002] At present, there are some auxiliary devices on the market to assist people to fall asleep, that is, sleep aids, so as to improve the sleep quality of users. Sleep state analysis is an important means for auxiliary devices to understand the user's sleep quality. During this process, the user's sleep state needs to be detected to accurately know whether the user is awake or asleep, and then corresponding intervention measures can be carried out. [0003] Polysomnography (PSG), also known as sleep EEG, is currently the "gold standard" for clinical sleep diagnosis and analysis. Polysomnography uses a variety of vital signs to analyze sleep. Among these signs, EEG is at the core; four rhythms of EEG are used: delta waves (1-3Hz), theta waves (4-7Hz), alpha Wav...

AI Technical Summary

Problems solved by technology

[0004] In general, to detect whether the user is awake or not is to use brain waves in four frequency bands (delta wave frequency band, theta wave frequency band, alpha wave frequency band and beta wave frequency band) to train the recognition model (classifier) ​​of the awake state. To identify EEG signals, these recognition models are often general recognition models that are trained using other people’s EEG signals, but due to the strong individual specificity of EEG signals, and the strength of EEG is very weak (EEG is microvolt level, the ECG is at the millivolt level), and it is very easy to be interfered by external signals during signal acquisition

[0005] In this process, when it is necessary to train a personal classifier, it is necessary to mark the data type of the collected personal EEG signal samples, so that self-learning and testing of the marked type of data can be carried out, and a more suitable personal classifier can be trained. When using a general recognition model to detect and mark EEG signals, as mentioned above, because the strength of EEG signals is very weak and easy to be interfered, it is easy to use a general recognition model to mark the types of EEG signal samples. Interference components are mixed in, resulting in low recognition accuracy of the trained personal classifier, which affects the reliability of the later detection results of personal sleep status

Images