ECG signal processing method based on two-level neural network with unbalanced training

Abstract

The invention discloses an ECG signal processing method based on an unbalanced training two-level neural network, which comprises the following steps: preprocessing: collecting ECG signal, adopting filter to eliminate baseline drift and power frequency interference, searching an R peak of the waveform of the ECG signal,and segmenting the beat of the ECG signal; signal recognition: ECG signals after heart beat segmentation are recognized by two-level neural network with unbalanced training, and abnormal ECG signals and normal ECG signals are obtained, wherein unbalanced training needs to be combined with two-level neural network; and compression processing: the ECG signal is subjected to adaptive compression based on intelligent diagnosis.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to an ECG signal processing method based on a two-level neural network of unbalanced training. Background technique [0002] Heart disease is one of the main diseases that threaten human life. For a long time, the research on heart disease has been an important topic in the medical field. Human electrocardiogram, as a comprehensive expression of heart electrical activity on the body surface, contains rich physiological and pathological information reflecting heart rhythm and its electrical conduction, so electrocardiogram is often used to analyze and judge various arrhythmias, and can also be used to diagnose myocardial damage It is of great reference value in guiding the treatment and rehabilitation of heart diseases, and it is also one of the most accurate methods for analyzing and identifying various arrhythmias. [0003] As an important routine examination method in c...

AI Technical Summary

Problems solved by technology

Its shortcomings are as follows: first, the ECG of the same patient at different stages is different, relying solely on feature engineering to extract these predetermined features may reduce the generalization ability of the entire algorithm; second, the process of extracting features is more complicated and may Increased computational complexity

Third, selecting appropriate features may be labor-intensive and time-consuming

However, the method of using neural networks has the following disadvantages: high processing power consumption, affecting the battery life of wearable devices, and occupying a large volume

In the compression process, there are the following deficiencies: First, the signal is completely transmitted through a simple lossless compression scheme. Although the original appearance of the signal can be well preserved, it is difficult to improve the compression ratio.

Second, if only a lossy compression scheme is used to transmit the ECG signal, although a high compression ratio is achieved, the quality of the compressed and recovered signal is poor, and important diagnostic information is easily lost

Images