Electrocardiogram signal classification and recognition method

Abstract

The invention discloses an electrocardiogram signal classification and recognition method. The method comprises the following implementation steps that original electrocardiogram waveform data with the measurement duration of 10 seconds or above is acquired, electrocardiogram rhythm information and a PQRST waveform are extracted according to the original electrocardiogram waveform data, and digitalized data of the electrocardiogram rhythm information and the PQRST waveform is acquired; a convolutional neural network is designed and constructed and trained, the acquired PQRST waveform data is input from the input end of the trained convolutional neural network, and type data is acquired through classification of the convolutional neural network. According to the method, by means of the fitting capacity of the convolutional neural network to a complex nonlinear function, more accurate and effective classification of ECG signals is acquired, so that high risk population, sub-healthy population and undetermined-condition population of cardiovascular diseases are monitored in real time, electrocardiographic changes in normal life, work and activities are intelligently analyzed to help determine condition or catch electrocardiogram information of potential cardiac diseases, and an early warning effect is made on a patient.

Description

technical field [0001] The invention relates to the field of medical technology, in particular to an identification method for classifying electrocardiographic signals. Background technique [0002] Hospital ECG examination, although the data accuracy is high, but can only record a segment of ECG waveform in a specific and very short period of time, for non-sustained arrhythmia, especially for transient arrhythmia and transient myocardial ischemic attack Often missed, delaying diagnosis. The 24-hour dynamic electrocardiogram (DCG) that is widely used at present can record long-term electrocardiogram (ECG) signals, but DCG has no ability to process data, cannot automatically classify signals, and cannot automatically identify pathological signals with medical significance. Wait for the 24-hour monitoring to end before the doctor can analyze the data and draw conclusions. [0003] Technologies for intelligent analysis of ECG signals are constantly evolving. By analyzing the...

AI Technical Summary

Problems solved by technology

However, the essence of wavelet transform is just a simple integral transform, and its nonlinear function fitting ability is very limited

Another existing technology uses support vector machine (SVM). SVM can be regarded as a 3-layer radial basis network with adaptive adjustment of network structure and weights, but it is limited in the ability to fit nonlinear functions. When given Training samples, the function family that SVM can fit is completely determined, but for more complex nonlinear functions, the current SVM is powerless

Images