Aimming at the problem of physiological and pathological characteristics and limitations of Fourier spectral analysis

put forward a method of Hilbert marginal energy radio as feature set under multiple frequency thresholds.Use the Hilbert Huang transform（ HHT）

get Hilbert marginal spectrum energy and make the the ratio of high frequency energy to total energy as feature. Compare to the energy spectrum of the Hilbert marginal difference of the normal and abnormal heart sounds

the range of frequency threshold is determined

and construct feature set

through the principal component analysis（ PCA） to reduce dimension

so as to remove redundancy feature

combined with support vector machine（ SVM） classification. 18 cases of normal people and 18 cases o

f patients are analyzed

and the results show that under the 80

160 Hz multiple frequency thresholds

the characteristic set has significant difference in the two classes of samples.