Human body heart and lung sound real-time blind separation method and system

Abstract

The present invention discloses a human body heart and lung sound real-time blind separation method and system. The method comprises: the step 1, collecting human heart and lung sound signals; the step 2, performing preprocessing of the human heart and lung sound signals, and obtaining human heart and lung sound signals after preprocessing; the step 3, converting the human heart and lung sound signals after preprocessing to a time-frequency domain, and obtaining a non-negative frequency spectrum; the step 4, employing the non-negative frequency spectrum to construct a blind signal matrix V, and obtaining initial matrixes Wk and Hk through separation of the blind signal matrix V, wherein Wk is a coefficient matrix, and the Hk is source matrix; and the step 5, collecting new human heart and lung sound signals, converting the human heart and lung sound signals to a new non-negative frequency spectrum after preprocessing, constructing a new blind signal matrix Vk+1 to perform online non-negative matrix decomposition, decomposing new matrixes Wk+1 and hk+1, and employing the matrixes Wk+1 and hk+1 to separate new human heart sound signals and lung breathing sound signals. The method realizes direct non-negative matrix decomposition for performing blind separation and does not need data recombination of legacy data and new data.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to a method and system for real-time blind separation of human heart and lung sounds. Background technique [0002] The current method of blindly separating heart and lung sounds using non-negative matrix decomposition, after collecting a piece of heart and lung sound signal, performs blind separation, and after separating the heart sound signal and lung breath sound signal, if a new heart and lung sound signal point is collected Finally, the previous heart and lung sound signals must be combined and then blindly separated again. [0003] For example, the hospital collects 100 heart and lung sound signal data points, and starts to separate the signals blindly to separate the heart sound signal and lung breath sound signal, and then collects a new heart and lung sound signal data point. Points are integrated together to become 101 data points, and blind signal separation i...

AI Technical Summary

Problems solved by technology

If a new heart and lung sound signal data point is collected again, the above steps must be restarted, and the old data and new data must be reorganized, which is cumbersome

[0004] The disadvantage of the existing method of blindly separating heart and lung sounds by using non-negative matrix decomposition is that it cannot be separated online. If the collected data points are 100,000 points, it takes 5 minutes to separate heart sounds and lung sounds. When new data points are obtained , it takes another 5 minutes to perform blind separation, and the old data and new data must be reorganized, which is cumbersome

Images