Research On Algorithm Of Characteristics Acquisition Of Pulse Waves And Its Applications

The accelerated tempo of life, unreasonable structure of food and aging of population are making the morbidity and mortality of Cardiovascular Disease increase gradually. People need a noninvasive method to examine cardiovascular disease. And this method can help them know their health status, find Cardiovascular Disease timely and ward off the accident for missing treatment opportunity. Pulse wave is one of the important physiological signals of body. Pulse wave analysis based on Traditional Chinese Medical Science provides an effective method to know the status of people's heart with non-ravage test. Now research on pulse wave analysis to get heart status is in primary phase. The relationship between pulse waves and typical pulse manifestations is not unambiguous and the applications of pulse wave analysis are limited. Thus, it is significant to study algorithms on characteristics acquisition of pulse waves and make pulse waves correlate with typical pulse manifestations both in theory and applications. The structure of this treatise is organized as below:Firstly, based on ideas at home and abroad, algorithms that are adopted to acquire characteristics of pulse waves can not include all useful characteristics of pulse waves. In this treatise different algorithms are adopted to acquire characteristics of different kinds of pulse waves: a mathematical model of pulse wave approximate by three-Gauss functions is issued and a hybrid GA is implemented to acquire characteristics of periodic pulse waves. Wavelet Transform (WT) is adopted to analyse aperiodic pulse waves and acquire its characteristics that are the energy of these kind of pulse waves wavelet transform in different scales.Secondly, since the background of pulse waves are complex and the classification basis of pulse waves are unambiguous, BP Neural Network (NN) is adopted to classify human pulse waves for its nonlinear mapping, and parallel computation ability. On idiographic algorithm design, appendix momentum method and auto learning rate adjust method are taken to mend the velocity of the algorithm convergent. In spite of limited training samples, the method in this thesis is superior to traditional pattern recognition methods if choose suitable features to be input-cell such as features extracted by algorithms designed above. The results also manifest that it is feasible to use Neural Network in classifying pulse waves.At last, the algorithms designed above are used in a system of pulse wave recognition. These algorithms are proved to be feasible and effective by data tests.As the algorithms implemented in this treatise are in accordance with characteristics of pulse waves and Chinese pulse manifestations. Hence, it could be used for reference in objective process of Traditional Chinese Medical Science and other pattern recognition systems.