A multimodal sleep staging method based on timing correlation drive

Abstract

The invention belongs to the technical field of sleep staging, and discloses a multimodal sleep staging method driven by timing correlation, which includes collecting polysomnograms and obtaining original signals for sleep state recognition; preprocessing the original signals to obtain sliced Fragment signals; feature extraction and fusion of segment signals to obtain feature vectors; time-series correlation embedding of feature sequences composed of feature vectors of multiple adjacent segment signals, and output sleep state prediction probabilities to obtain sample prediction results; in model training stage, calculate the loss value between the prediction result corresponding to the training sample and the real sleep state of the training sample, and train the sleep staging model to obtain the trained sleep staging model; in the model evaluation stage, input the sample to be evaluated into the trained The sleep staging model obtained the prediction results corresponding to the samples to be evaluated. The invention solves the problem of multi-modal signal feature extraction fusion and timing correlation.

Description

technical field [0001] The present invention relates to the technical field of sleep staging, in particular to a multi-modal sleep staging method driven by timing correlation. Background technique [0002] People will experience periodic changes of different sleep states during the sleep process. The task of sleep staging is to judge the sleep state of people through the physiological signals (such as polysomnography, PSG) recorded during the sleep process. As an important reference for the diagnosis of sleep disorders. Traditional sleep staging research is usually based on the technical path of feature engineering. The advantages of feature engineering-based solutions are low training complexity and good model interpretability. The disadvantage is that feature extraction is strongly dependent on prior knowledge. With the accumulation of a large amount of labeled data, the substantial growth of computer performance, and the breakthrough of deep learning model training algor...

AI Technical Summary

Problems solved by technology

First, the existing models do not fully consider the heterogeneity among modalities (EEG, EEG, etc.) in sleep physiological data; second, the evolution rules of sleep states, that is, the temporal correlation of sleep states, are not reasonably applied to the model middle

Images