Research About Information Processing Methods In Pulse Diagnosis Objectivization

Traditional Chinese Medicine (TCM) diagnosis acquires pathological information through observing, smelling, asking and touching, which are also called "Four Diagnosis". As a major part of touching, pulse diagnosis becomes an important method to find out health status of patients. However, TCM, lack of quantitative standard, excessively relies on doctor's subjective judgment. So, objectivizing of TCM tries to establish diagnosis standard in order to solve these problems.This paper summarizes from both success and defeat of our predecessors in this field. Based on as TCM principle, it utilizes advanced digital signal processing technique to make further research into information processing technique of Pulse Diagnosis Objectivization.Mainly according to diagnosing experience, traditional pulse diagnosis is lack of objective standard, inconsistently, computer-based pulse diagnosis device, however, depends on the quantitative principles. provides a good solution for this inconsistency and makes TCM Objectivization possible and feasible. It not only assimilates integral viewpoint of TCM, but also makes quantitative diagnosis by mathematic method. More importantly, develops TCM theory; meanwhile, it makes a radical difference comparing to TCM.Concerning about pulse signal collection, we apply modern Measurement Technique and signal processing theory, design unique pulse automatic and layered measuring system. In order to collect digital signal in real-time, this system utilizes PCI data collecting card to transform analog signals into digital signals, and then data processing program to carry out signal analysing and displaying.In the aspect of pulse signal pre-processing, this paper proposes a new EMD (Empirical Module Decomposition) based filter to remove baseline wander. Additionally, we made simulated experiments and evaluated its validity by quantitatively comparing with morphology filter, time variant filter and spline Estimation. The results showed that the proposed algorithm is efficient in removing pulse baseline wander.In the aspect of pulse characteristics extraction, we not only make comprehensive analysis from time and frequency field, but also creatively apply EMD, Hibert-Huang Transform and Cepstrum Analysis into pulse characteristics extraction. We can find out that these parameters are efficient in the following classification process.Following the pulse characteristics extraction above, we choose 9 characteristics parameters to construct characteristics vector, and utilize improved Binomial Tree based SVM (Support Vector Machine) to execute classification of 7 typical kinds of pulse. Finally, we compare this algorithm with DAG (Directed Acyclic Graphs) SVM and Neural Network, and we find that right identification rate increases about 3-6 percent.