Research On Pulse Feature Extraction,Classification And Recognition Based On Deep Learning

Pulse signal is a kind of weak non-stationary signal,which is the phenomenon of the pulse beating felt by the fingers,which can help us to find out which part of our body develops the disease and judge the nature of the disease as well as the severity of the condition.It has a great significance to judge and identify the disease,because lots of diseases,physiological phenomena and the function of internal organs(visceral functions)can be reflected thro-ugh the pulse.However,in Traditional Chinese Medicine pulse diagnosis,physicians often make diagnoses based on their subjective feelings and long-term diagnostic experience.Therefore,there is no objective basis for diagnosis in traditional Chinese medicine.In order to avoid the influence of artificial and subjective factors on the diagnostic results of pulse images,this paper explores and discusses the process of digitizing and recognizing choroidal and synovial pulse images,which are commonly found in clinic.We unfold the study on the identification of pulse analysis based on deep learning algorithm,which can automatically learn the characteristic information of the pulse signal and thus realize the objective analysis identification and data transmission on the choroidal and slippery pulse images.It provides some guidance for the objectification and intellectualization of pulse diagnosis.This paper designs a traditional Chinese medicine pulse recognition system based on Python language,which mainly includes pulse analysis and recognition part and data transmission part for transmitting recognition results to the server.The pulse analysis and recognition part includes the recognition model based on mathematical method and the recognition model based on deep learning method.In the mathematical method recognition model,firstly,the basic structure and significance of pulse map are analyzed,then the pulse characteristics and clinical significance of string pulse and smooth pulse are introduced,and the judgment criteria of two kinds of pulse are given,Then,after preprocessing the pulse wave data,the pulse characteristics are extracted by time-domain analysis method for pulse recognition from the perspective of mathematics,and the correctness of the recognition results is analyzed;In the deep learning method recognition model,firstly,the pulse wave data points are transformed into time domain diagram and frequency domain diagram,and then a pulse recognition model based on mobilenetv3 neural network framework is established.Adam and dropout optimization algorithms are introduced to supervise the learning of the two kinds of pulse,and the recognition results are compared with the mathematical method.After practical testing,the traditional Chinese medicine pulse system designed in this paper can complete the reception and processing of pulse signals,realize the recognition and result transmission of string pulse and smooth pulse,and provide a certain guiding significance for the objectification of pulse diagnosis.