An ECG signal detection device and analysis method based on joint neural network

Abstract

The invention discloses an ECG signal detection device and analysis method based on a joint neural network. First, a joint neural network algorithm is built on a machine learning server and a model is trained. For the preprocessed ECG data, the model is passed through the residual neural network. The network module extracts the spatial features of the data and obtains the spatial classification probability, and extracts the time-series features of the data on the dimensionality-reduced spatial feature map through the bidirectional long-short-term memory neural network and the attention module to obtain the time-series classification probability, and finally fuses the two classification probabilities Obtain the test results; obtain a small amount of ECG data from the wearable device, manually mark it and input it into the machine learning server for model fine-tuning, and deploy the final model to the smart mobile device; finally, the wearable device and the smart mobile device will be connected through wireless Transport enables real-time anomaly detection. The present invention develops a wearable device from collection to real-time detection of electrocardiographic signals, and provides effective technical means for assisting in the diagnosis of heart diseases.

Description

technical field [0001] The invention belongs to the field of abnormal detection of electrocardiographic signals, in particular to a wearable device for electrocardiographic signals based on a joint neural network algorithm and an abnormal detection and analysis method. Background technique [0002] The ECG signal is drawn by the electric current generated by the heart activity and transmitted to the body surface. It can accurately reflect the activity state of the heart and is one of the important means for doctors to diagnose heart disease. ECG signal is a weak physiological signal with small voltage value and time interval. Therefore, extracting ECG signal features and assisting detection by computer can improve the efficiency of doctors' diagnosis. In recent years, with the continuous development and improvement of artificial intelligence algorithms, algorithms commonly used for ECG signal detection and classification are mainly divided into traditional machine learning m...

AI Technical Summary

Problems solved by technology

Many studies only use one type of neural network when building models, resulting in the inability to make full use of the characteristics of ECG

And although some researchers use CNN and LSTM models at the same time, the superimposed structure will cause the neural network algorithm to allocate most of the error to the LSTM network that is closer to the classification layer when the neural network algorithm updates parameters through backpropagation, resulting in a shallow extraction space. The CNN network cannot be updated to the maximum extent

In addition, many researchers did not consider the issue of patient specificity when training and testing algorithms, that is, the training set and test set data they used may come from the same patient, which will lead to poor generalization ability of the final model. ECG data of new patients cannot be accurately detected

Images