Research On Recognition Method Of Cardiovascular Disease Based On Pulse Wave Signal

Globally,the morbidity and mortality of cardiovascular diseases are increasing year by year,seriously threatening people’s life and health.Cardiovascular disease is a disease that can be controlled to a certain extent through prevention.Early detection and early treatment of cardiovascular disease can fundamentally reduce the harm of cardiovascular disease and effectively identify early cardiovascular disease has great significance.The pulse wave contains important physiological and pathological information of the human body and can reflect the overall function of the cardiovascular system.With the advancement of the objectification of pulse diagnosis in traditional Chinese medicine and the continuous development of artificial intelligence,more and more researches have been done on the classification of pulse waves using neural networks.The identification of cardiovascular diseases based on pulse waves is a non-invasive and simple inspection method that can be detected anytime and anywhere,which is helpful for early detection of abnormalities,timely intervention and treatment,and can effectively delay the progress of the cardiovascular event chain.This paper studies a deep learning algorithm for identifying patients with cardiovascular disease and normal people based on fingertip pulse waves,and mainly completes the following tasks:(1)Collected 283 fingertip pulse wave data of patients with cardiovascular disease and 255 fingertip pulse wave data of normal people,and preprocessed the collected data to remove noise interference,improved the quality and reliability of the signal,and formed a pulse wave Wave classification dataset.(2)The ResNet network model and MobileNetV3 network model suitable for pulse signals were built,and the model structure and network parameters were adjusted through a large number of comparative experiments to continuously optimize the model and improve model performance.(3)The data set obtained after preprocessing is used as the input of the neural network model,trained through the ResNet network model and the MobileNetV3 network model,and the accuracy rate,loss function,parameter amount and calculation amount of the two models are compared and analyzed.The experimental results show that the two models have achieved an accuracy rate of more than 80%in the validation set and can classify and identify patients with cardiovascular disease with an accuracy rate of more than 80%in the test set,and achieve 100%accuracy in the test set.For the classification and recognition of normal people,the accuracy rate of the overall test set is above 90%.While the MobileNetV3 network has far less training parameters and calculations than the ResNet network,the accuracy rate is the same as the ResNet network,and the model training process is much faster,saving a lot of time and computing resources.