Research On Denoising And Detection Of Pulse Signal Based On Stacked Autoencoder

Heart rate is one of the most important physiological parameters of the human body and plays a large role in clinical diagnosis and patient health monitoring.Abnormal heart rate is a high-risk factor for many cardiovascular diseases.By monitoring heart rate,many cardiovascular diseases can be prevented in advance.At present,the most accurate methods for measuring heart rate are electrocardiogram and pulse oximetry sensor.The electrocardiograph with high accuracy but inconvenient to use in the hospital can no longer meet the needs of people for daily heart monitoring.Therefore,wearable heart rate measurement devices based on photoplethysmography(PPG)technology are widely used in the field of heart rate monitoring.The wearable heart rate detection system based on PPG has low cost and is widely used,but it is susceptible to interference and the measurement accuracy needs to be improved.The PPG signal acquisition process often contains signals such as power frequency interference,baseline drift,myoelectric noise,etc.In view of the shortcomings of the current heart rate monitoring method with low accuracy of heart rate monitoring,a deep learning algorithm is proposed to extract the heart rate value in PPG signal.The stack noise reduction self-encoding network is an unsupervised deep learning model.It is formed by stacking several noise reduction self-encoders through a layer-by-layer stacking mechanism similar to deep networks.In this paper,the research on the application of stacked autoencoder to the denoising and detection of pulse signal is carried out.The first part is the denoising and heart rate measurement of pulse signal,and the other part is the research of abnormal heart rate detection.The main contents of this article are as follows:(1)PPG signal denoising,how to eliminate the effect of motion noise in PPG signal on reliable heart rate measurement.At the same time,the PPG signal measures the heart rate.In the detection of physiological parameters,the heart rate can be used to monitor whether the daily exercise amount of the human body exceeds the standard,and can also provide a reference for medical diagnosis.Compared with the Fourier transform and the wavelet transform,this paper is based on the denoising of the pulse signal of the stacked autoencoder,So as to realize the heart rate extraction of the PPG signal with severe interference in the exercise state,combined with the adaptive threshold(ADT)method to calculate the heart rate.(2)PPG signal calculates the heart rate for heart rate abnormality detection.Each tester’s PPG signal corresponds to its unique features,and deep learning learns the inherent laws of these features.This paper uses deep learning convolutional neural network classification method to distinguish the diagnosis of abnormal or normal heart rate.In view of the difficulty in learning features in traditional arrhythmia intelligent diagnosis and the need to master a large number of signal processing methods anddiagnostic experience,a new method for classifying and identifying arrhythmia states directly from PPG signal data is proposed.This method eliminates the explicit feature extraction stage of intelligent diagnosis,which can reduce the factors of manual participation and get rid of the dependence on a large number of signal processing techniques and diagnostic experience.The method studied in the experiment has good recognition ability for the realization of heart rate anomaly detection,has a good recognition ability,can complete the adaptive extraction of arrhythmia features,and enhances the intelligence of medical heart rate signal denoising and abnormal detection.