Research On The Extraction And Recognition Technology Of The Pulse Time-domain And Frequency-domain Features

Currently, cardiovascular disease has been the primary killer who causes harm to the health of people on a global. Most methods used in the clinical diagnosis are invasive, not only does it bring great pains to patients, but also real-time monitoring can't be realized. Non-invasive diagnostic method has developed for many years and also achieved a great success. Scholars use information about physiological processes just as the relation between the parameter of pulse wave and the coronary heart disease pathological changes. Scientific research personnel summarize a great deal of parameters which are about the characters of pulse in both time and frequency domain. But the number of parameters is very enormous, and lack of an effective evaluation standard to appraise the correlative degree between the parameters and cardiovascular disease pathology characters; the relationship between the individual parameter and physiological information is ambiguity, lack of systematicness, so now the non-invasive diagnostic method only can be used to appraise the individual item but not diagnose the disease. All these problems hold it back to apply in the daily clinic health-care. And then, considering the excessive duplication of the pulse parameters research in time and frequency domain, it's imperative to creating new research idea under the situation.This paper focuses on the problems about that pulse wave research used in diagnosis of cardio vascular disease with the analyses on pulse wave time and frequency domain. Searching the typical features of the coronary heart disease from many characters parameters by feature extraction technique, and resorting these parameters to build the recognition model, we make the relationship between the simple parameter list and disease diagnosis. This paper pushes the complexity parameter—method of wavelet entropy analysis to pulse character identifies frontiers, creating new research idea of pulse wave.Pulse waves of 30 normal subjects and coronary heart disease subjects are picked up respectively. All the parameters above in time and frequency domain are both computed, and feature extraction is used by genetic algorithm. So we get rid of the parameters which are not irrelevant with the diagnosis of coronary heart disease and find out the striking relevant parameters to achieve the pick-up of the Characteristic parameters of coronary heart disease. On the base of the method of optimization this paper analyses the pulse wave parameters in time domain and build the model in the method of fuzzy pattern recognition, and then take the cases at random to check the model. The result shows the recognize model can achieve the test precision 100%. We build the model of pulse frequency domain with the method of Bayes discriminant analysis, and then take the method of cross certification to check the model. The result indicates the recognition model can achieve the test precision 90.0%,83.3%and 80.0%,83.3%. It suggests that this paper achieves the goal to diagnose coronary heart disease by pulse character system well. At the end of the paper, we compare two kinds of model building methods and validate methods in order to make the preparation for future research.