Application Of Hilbert-huang Transform To Human Pulse Signal Analysis

Pulse signal is one of the most primary physiological signals, including the pulse frequency, the pulse rhythm, pulse-filling ratio, state of stream, state of motion, fluctuation amplitude and so on. Formation of pulse signal has close relations with the blood and viscera. Therefore, the pulse signal can be reflected the change of mechanisms and pathophysiology in the blood and viscera. So we can use modern signal processing methods to analyze pulse signals of healthy normal peoples and heroin addicts, and to find significant different characteristic parameters which can classify the pulse signal of healthy normal people and heroin addict.Have compared with some other methods used previously, a new one suitable for nonlinear and non-stationary digital signal sequence called HHT method is adopted to deal with the pulse signals. By analyzing the pulse signals of healthy normal peoples and heroin addicts through the HHT method, different features can be fetched.Fetching feature is very important in this paper. This paper gets three types of features.1. Analyzing Hilbert power spectra of pulse signals of one healthy normal peoples and 15 heroin addicts by statistical analysis technique, Power-Spectral Energy Ratio can be got as a feature; 2. Comparing intrinsic mode function components of pulse signals, there is a difference in the amplitude of the last IMF component. So this amplitude can be got as a feature. 3. Comparing Hilbert bound spectrum of pulse signals, the heroin addict significantly raised pulse signal in the 3-5Hz range, so energy ratio of power in this range and total power can be got as a feature. Method 1: selecting energy ratio in 3-5Hz range and Power-Spectral Energy Ratio in 0-5Hz range to form a two-dimensional feature vector, this vector can be used to classify the pulse signal of healthy normal peoples and heroin addicts. The overall identification success rate is 90%, so the desired result can be achieved. Method 2: selecting energy ratio in 3-5Hz and the amplitude of the last IMF component to form a two-dimensional feature vector, this vector can be used to classify the pulse signal of healthy normal peoples and heroin addicts. The overall identification success rate is 90%, desired result can be achieved, too.At last, in the basis of the feature extraction of pulse signals, this paper also uses the Support Vector network to identify the 30 pulse signals. (15 healthy normal peoples, 15 heroin addicts).The results show that these two methods can achieve the desired results. Therefore, the HHT analysis method is effective method of pulse signal characteristics extracting.