Method and system for labeling type of electroencephalographic signal data in sleep state

Abstract

The invention relates to a method and system for labeling the type of electroencephalographic signal data in a sleep state. The method includes the steps that wavelet decomposition is carried out on an electroencephalographic signal sample, an electroencephalographic signal is reconstructed according to a preset wavelet coefficient in a low-frequency stage, and a low-frequency electroencephalographic signal is obtained; brain waves are extracted from the reconstructed low-frequency electroencephalographic signal; K synthetic waves and delta waves are detected from the brain waves according to the waveform characteristics of the K synthetic waves and the delta waves; the number of the detected K synthetic waves and delta waves is calculated, and when the number exceeds a preset number threshold value, the signal type of the electroencephalographic signal sample is labeled as sleep. By labeling the type of the electroencephalographic signal sample according to the technical scheme, interference elements mixed into the electroencephalographic signal sample can be eliminated, so that a personal classifier trained with the labeled electroencephalographic signal sample is higher in recognition accuracy rate, and the reliability of the detection result of the personal sleep state detected in a later stage is 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 EEG signals in a sleep state. Background technique [0002] There are already some devices on the market to help people fall asleep and improve the quality of sleep. Sleep state analysis is an important means to understand the user's sleep quality. 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 Wave (8-12Hz), frequency characteristics of beta wave (14-30Hz). Since the strength of the EEG signal is very weak (the EEG signal is at the microvolt level, and the ECG signal is at the millivolt level), it is very easy to be interfered by external signals during signal collection and...

AI Technical Summary

Problems solved by technology

[0003] At present, the EEG signal data for detecting sleep state is generally recognized by training the recognition model, that is, the EEG signal collected by other people is used to pre-train a classifier (also called a general classifier). Among them, 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 can be performed on the marked type of data, and a personal classification that is more suitable for individuals can be trained However, when using a general-purpose classifier to label data, as mentioned above, because the strength of the EEG signal is very weak, using this method to label the type of the EEG signal is easy to mix in interference components, resulting in the training of individuals. The recognition accuracy of the classifier is low, which affects the reliability of the individual sleep state detection results in the later stage

Images