A segmentation and localization method of heart sounds based on vmd and multiwavelets

Abstract

The invention discloses a heart sound segmentation and localization method based on VMD and multi-wavelets. First, the center sound interval and type of heart sound signal samples are marked, the heart sound signal samples are segmented by sliding windows, and the kurtosis is used to decompose each of the samples obtained by VMD. IMF components are selected, the most effective IMF components are decomposed by GHM multi-wavelet packet, the time-frequency domain matrix of heart sound signal samples is constructed based on the decomposition results, and the sub-matrix is ​​extracted from the marked heart sound interval and the time-frequency column vector is obtained, which is used as The training samples are used to train the preset classification model; the same method is used to obtain the time-frequency domain matrix of the heart sound signal to be segmented and localized, the Shannon energy envelope is extracted, the heart sound interval is determined based on the maximum inter-class variance method, and the sub-matrix is ​​extracted and obtained by the same method. The time frequency column vector is input into the trained classification model to obtain the segmentation and localization result of the heart sound signal. The present invention can improve the performance of heart sound segmentation under high noise.

Description

technical field [0001] The invention belongs to the technical field of heart sound processing, and more particularly relates to a heart sound segmentation and localization method based on VMD and multi-wavelets. Background technique [0002] At present, the use of wearable heart sound detection equipment for heart sound monitoring has become an important means of diagnosing and controlling heart diseases. Considering the complex environment in which the wearable heart sound detection device is used, other physiological sound interference, background noise, power frequency interference, and motion artifacts may be introduced during use, which will seriously affect the further processing of the collected heart sound signals. [0003] The traditional heart sound segmentation and localization algorithm is combined with the electrocardiogram. When the electrocardiogram is used as a reference, the effect of heart sound segmentation is very good. However, the ECG, as another signa...

AI Technical Summary

Problems solved by technology

The derivation of parameters based on the envelope extraction method is directly affected by the threshold, and the selection of the threshold mainly relies on empirical values, so it is easy to cause poor generalization between different sample sets due to over-learning

[0008] Generally speaking, although the current method can realize the difficulty of identifying S1 / S2 in normal heart sound signals, there are still many problems if you want to use traditional heart sound processing methods on the wearable platform: the wearable platform faces Scenarios with complex noises, such as murmurs, clicks, crackles, and low signal-to-noise ratios, will greatly increase the difficulty of heart sound segmentation and positioning

Images