Research On Calibration And Standardization Of Pulse Taking Instrument

The first step of objectification of pulse diagnosis is developing reliable pulse acquisition instrument which can gather objective and accurate data. However, pulse is variability as one kind of dynamic physiological signal. It is a research emphases that if the pulse taking instrument can capture pulse signal repeatedly. The aim of this dissertation is to test reproducibility and objectivity of the optimized device for data collection.This dissertation optimizes the signal processing circuit of existing device and improves signal quality and intensity base on noise analysis. It modifies the circuit to modularity on the premise of keeping the input and output unchanged which could reduce the volume of the device and makes maintenance easier. It also adapts the bottom level driver and PC client for making data transfer more safety and easier using.This dissertation chooses wavelet pocket to resolve signals and select threshold according the characteristic of noise. Minimum entropy could compute the optimal wavelet packet basis which cancel 50 HZ noise effectively and reserve details information. Baseline drifts removing uses cubic spline interpolation and could choose different interpolation points with different demand. The dissertation improves the method of period segmentation and makes it simple and preciseness.Before similarity measurement, this dissertation extracts signal features for better signal representation and reducing processing complexity. The extracted features include 12 dimensions of temporal properties and 5 dimensions of spatial properties(the energy features).Standardization research uses statistical methods, two-sample t-test and one-way analysis of variance which are often using in medical analysis. In order to obtain convincing results both in medical and computer science fields, this dissertation uses dynamic time warping to select similar cycles in with-in sample. Although statistical methods have anti-interference ability partly, they are helpless for reflecting real similarity when noises up to a certain extent. DTW could remove abnormal periods and choose the more similar periods to extract feature.With the optimized features by DTW, statistical methods can draw a scientific and objective conclusion that the signals gather from same people by different devices are similar.