The Nonlinear Dynamical Characterization Analysis Of The Human Physiological Parameters

Hypertens ion，coronary heart disease and other cardiovascular disease are thema in indicators of phys iologica l activity. It is possible to warn some diseaseearly and reduce the inc idence of une xpected situatio ns if we can find theirchanges in time.This paper studies three physio logica l parameters of human (ECG, respirationand body te mperature), and focuses on ECG feature detection and ana lys is. First,we introduce the AHA database, MIT-BIH database and some basic knowledge ofthe ECG. Secondly, we introduce three common nonlinear identification method,includ ing permutation entropy, Lyapunov exponent and K olmogorov entropy, andthen apply these methods to three groups of ECG (norma l signa l, atria lfibr illatio n signa l and ventr icular fibr illation signa l). According to theexper imenta l results, the perfor mance of permutation entropy for classification isthe best while Kolmo gorov entropy is poor. For a ll these methods, theperfor mance of the c lassification o f the nor ma l signal a nd ventr icular fibrillationis quite good, but the classification results of the norma l signa l and AF signal arenot good. To address the reason, we find that the difference of the nonlineardyna mical characteristics between the norma l signa l and the AF signal is notobvious. There fore, we focus on the AF signa l detection algor ithm so as to obtainbetter AF classification precis ion. Thereby, we combine the RR interva l andShannon entropy to detect AF signa l. We first realize the QRS detectionalgor ithms, and then use the preprocessed RR interva l var iance and Shannonentropy to detect the AF signa l. We va lidate the algorithm with MIT-BIHdatabase. According to the results we can observe that the hybr id algor ithmoutperforms any previous single algorithm.Fina lly, this paper also emphas izes other typ ical phys iologica l time series(respiratory signals). We consider the brute-force algor ithm to study thesynchronization of ECG and respiratory signa ls. We emp loy a period of stab lerespiratory signa ls which is then coupled together with the diverse ECG signa ls,includ ing the norma l, atria l fibr illatio n and ventr icular fibr illatio n signa ls. Bydetecting the deviation a mong each subsequence of the new signa l, we identify the sync hronizatio n relations hip between the ECG and respiratio n signa ls. Weconc lude that the brute-force algor ithm is quite effic ient on the synchro nizationof huma n bio medica l signa ls. In additio n, we design a new method to quick lydetect the body temperature as it a lways takes a long time to measure bodytemperature by conventiona l methods. We pre-heat the temperature sensor to35degree and then make curve fitting to the ten-second temperature data to find thetemperature convergence. So our approach realizes the detection of bodytemperature in less than30second.In summary, we have basica lly comp leted the characterization ana lys is of thehuma n phys io logica l time series and validate the rationa lity and effectiveness ofthese algor ithms. These analys is constitutes the basic functio nal co mponents ofthe ECG expert system and achieves the require ments of potentia l engineerapplication.