Research On Multimodal Heart Function Signal Analysis Method

Cardiovascular disease has a great impact on human’s health.Cardiac function assessment plays an important role in the whole cycle of cardiovascular disease examination.Both cardiac impedance signal and electrocardiogram signal can objectively reflect the condition of human’s heart.Cardiac impedance differential signal carries a lot of cardiac physiological and pathological information,which can accurately reflect the hemodynamic changes of the heart in real time,evaluate the cardiovascular function,and contain more comprehensive information features after fusion with electrocardiogram signal.Therefore,compared with the single analysis of a single heart function signal,the fusion analysis of cardiac impedance and electrocardiogram signal collected at the same time will get more comprehensive information and better understand the state of the heart.In this paper,combined with signal processing technology and deep learning technology,a multi-modal heart function signal classification model based on multi-scale feature fusion is proposed to realize the automatic evaluation of human heart function.According to the generation mechanism of physiological signal of cardiac function,the experimental scheme of physiological signal acquisition of cardiac function is designed.The multi-modal cardiac function signals of four different exercise intensities are collected synchronously,and two single-modal physiological signal databases are constructed respectively.Combined with the corresponding relationship between the two single-mode cardiac function signals,the wavelet transform method and threshold method are used to preprocess the two single-mode cardiac function signals,such as de-noising,feature point location and cardiac cycle division.Because there are many important physiological parameters about human’s body in electrocardiogram signal and impedance signal,the two signals describe the function of heart from different angles,which provides a reliable basis for diagnosis.This paper proposes a classification model of cardiac function signal based on multi-scale feature fusion,which can automatically extract the deep features of cardiac impedance differential signal and electrocardiogram signal,reduce some subjective effects caused by manual feature extraction,and obtain the classification and evaluation results of the cardiac function signals under different exercise intensities.In order to compare the classification performance of the model when multimodal and single-mode heart function signals are input,the feature extraction and fusion of multimodal heart function signals are carried out by using the designed deep neural network model,the internal relationship between the signals is mined and learned,and the heart function signals under different motion intensity are automatically classified.The results of the classification of single mode cardiac function signals and multimodal cardiac function signals by comparing the designed network show that the analysis based on multimodal cardiac function signals can improve the average accuracy rate in the test set,reaching 97.28%,and the probability of missed and misjudged of the model is low,which can describe the hemodynamic changes of heart more accurately,evaluate cardiovascular function,and check the heart related diseases is very important to check,prevent and treat early.