Feature Extraction of Pulse Signal Based on Hilbert-Huang Transformation and Singular Value Decomposition

A pulse diagnosis approach based on Hilbert-Huang transformation method and Singular Value Decomposition (SVD) technique is proposed. The Empirical Mode Decomposition (EMD) method is used to decompose the signal into a number of IMF components, then applying the Hilbert transformation to creating analytic signal and obtaining instantaneous frequency and instantaneous amplitude, from which the initial feature vector matrices are formed. By applying the singular value decomposition technique to the initial feature vector matrices, the singular values are obtained, which are regarded as the state feature vectors of the human pulse signals. Finally the first 20 singular values of SVD are showed in a parallel coordinate's graphic form. Practical examples show that the proposed approach can be applied to pulse diagnosis effectively. A method of mining pulse signal is presented for extracting the time-frequency distribution feature of the data based on the technique of the singular value decomposition. By the time-frequency analysis, the important pulse characteristic information is extracted, the research provide the basis for further classification. This provides one new method for the pulse diagnosis thorough research. It will be helpful to make the objectification of pulse study.