Electroencephalogram signal online prediction method based on kernel recursive least square self-adaptive tracking algorithm

Abstract

The invention relates to an electroencephalogram signal online prediction method based on a kernel recursive least square self-adaptive tracking algorithm. The electroencephalogram signal online prediction method is used for solving the problems that electroencephalogram signals detected online are inaccurate and missing. The method includes a training stage and a prediction stage, and the training stage comprises: generating electroencephalogram signal labels; selecting a forgetting strategy, and updating the intermediate parameters; performing online prediction on the electroencephalogram signal to obtain a noise-free variance, a predicted average output signal segment and a predicted variance; calculating a root mean square error of a predicted output signal fragment of the current iteration and an output electroencephalogram time sequence fragment label, adjusting intermediate parameters, and if the root mean square error of all fragments in all samples reaches the minimum, completing training, and performing an online prediction stage; and otherwise, continuing training. According to the method, the electroencephalogram signal prediction is realized, and the problems of inaccurate electroencephalogram signal measurement and signal loss are solved.

Description

technical field [0001] The invention relates to the fields of brain electric signal preprocessing and biological interdisciplinary medicine. Background technique [0002] The purpose of modeling and analyzing the time series of EEG signals is to discover and extract valuable information in EEG data, and to reveal the inherent implicit connections between EEG signals. EEG time series prediction plays an important role in the time series analysis of EEG signals. [0003] In terms of time-series analysis of EEG signals, first, Bag-of-wave features are used to learn EEG synchronization patterns to predict epileptic seizures. Second, machine learning methods as well as classic deep learning methods such as CNN are applied to the prediction of seizures from EEG signals. Furthermore, a method combining co-spatial patterns (CSP) and convolutional neural networks (CNNs) was used to predict seizures from EEG signals. Seizure prediction from EEG signals can improve quality of life b...

AI Technical Summary

Problems solved by technology

[0005] The invention is used to solve the problem of inaccuracy and lack of online detected EEG signals, creatively introduces the kernel recursive least squares adaptive tracking algorithm into the field of EEG signal prediction, and predicts the future through the measured historical data of EEG signals EEG signal, realizes the prediction of EEG signal, solves the problem of inaccurate EEG signal measurement and missing signal, and lays the foundation for the next step of disease diagnosis

Images