Design Of A Portable Heart Sound Signal Acquisition Instrument And Research On Classification Algorithm

Cardiovascular disease is one of the most important diseases that threaten human life and health safety,and the improvement of prevention and diagnosis of cardiovascular disease is of great significance to reduce its mortality rate.However,most hospital tests are not portable and expensive because of the uneven distribution and limited number of medical resources in most areas of China,so patients are often unable to detect abnormalities in the early stages of the disease,and usually go to the hospital for testing only after they experience abnormal heart function and discomfort,which delays the best treatment time and leads to the aggravation of the disease.Heart sound signals contain rich physiological information about the heart and cardiovascular system,and intelligent diagnosis of cardiovascular diseases based on heart sound signals and deep learning algorithms has received much attention from researchers at home and abroad,but there are not many portable heart sound auscultation instruments that integrate intelligent diagnostic functions.To address the above-mentioned problems of portability and intelligence of existing medical devices,a portable heart sound smart detector device,based on the heart sound signal obtained from the MEMS high-precision heart sound sensing chip developed in the laboratory,is developed in this paper.Firstly,the hardware circuit is designed to filter,amplify and transmit the obtained heart sound signal,and then the hardware and software processing program is designed to further process the heart sound signal for display and playback and so on.Finally,combine the deep learning heart sound classification algorithm and cloud server to build a portable heart sound intelligent detection instrument.The main work of this paper includes the design of a portable heart sound signal acquisition instrument,the study of deep learning heart sound classification algorithm,and the construction of a cloud-based intelligent diagnosis system.(1)Design of portable heart sound signal acquisition instrumentBased on the high-precision MEMS heart sound sensor,this paper designs the supporting signal processing and other hardware circuits,develops the device side,cell phone side application and software database cloud server platform,develops a portable heart sound detection instrument,and builds a heart sound signal processing system integrating heart sound signal acquisition,transmission,display and playback.(2)Deep learning heart sound classification algorithm researchThis paper extracts two-dimensional features based on the original one-dimensional heart sound data,and extracts four different features including envelope and log-Mel spectrogram after considering the feature domain and clinical significance,and selects classical neural network algorithms such as ResNet50 and MobileNetV2 by comparing the functional structure and feature parameters of the networks for comparison experiments.Based on the experimental results,the most applicable combination of MobileNetV2 and log-Mel spectrogram was selected to build a deep learning diagnostic model and deployed in the cloud server to enrich the intelligence of the heart sound detector.After that,the comparison of the effects of different features is analyzed and discussed,and the objective reasons for the different effects achieved by different features are systematically analyzed.Finally,the accuracy of the experimental results is verified using transfer learning algorithms and traditional machine learning algorithms and the impact of using transfer learning on the experimental results is investigated.(3)Cloud-based intelligent diagnosis system constructionBased on the hardware and software design of the instrument and the construction of the heart sound classification algorithm,the software backend database and the classification algorithm are deployed in the cloud server to build a cloud-based intelligent diagnosis system integrating heart sound signal processing and display,cloud-based electronic medical record and intelligent diagnosis algorithm.