Research On Key Technologies Of Lung Sound Signal Processing Based On EMD Technology

EMD(Empirical Mode Decomposition)method is an adaptive time-frequency analysis method based on signal local features for non-stationary and nonlinear signals proposed by American Chinese N.E.Huang in 1998.Compared with time-frequency methods such as Fourier analysis and wavelet transform based on the tradition linear and stationary assumptions,EMD algorithm has a unique advantage in dealing with non-stationary nonlinear signals.It has important research value in theoretical research and engineering applications.Lung sound signal processing is an important means of non-destructive detection of human lung diseases.In today’s respiratory disease has become a threat to human health.The use of computer technology to record,detect,identify,quantitatively analyze and assist the diagnosis of pulmonary sound signals is of great significance not only to improve the accuracy of the actual clinical diagnosis,but also to the related theoretical research.In this paper,further research on the EMD algorithm is made.From the theoretical point of view,the problem of envelope fitting and end effect in the EMD algorithm is studied,and improved methods is proposed.Then from the application point of view,for the lung noise denoising problem,the EMD denoising algorithm is deeply studied and applied to the lung noise denoising.A new method of lung noise denoising based on EMD and blind source separation is proposed.The main contents and innovations of this paper are as follows:(1)For the envelope overshoot or undershoot and mode mixing problems caused by traditional cubic spline interpolation in EMD algorithm,a rational four-time Hermite interpolation algorithm: ORHEI(Optimal Rational Hermite Envelope Interpolation)is introduced to flexibly change the envelope by adjusting the shape parameters.In order to achieve the optimal effect,the PSO algorithm(Particle Swarm Optimization)is used to find the optimal curve in the curve cluster to complete the optimization and selection of the envelope curve.The effectiveness of the algorithm is verified by the simulation of the signal and the actual nonlinear non-stationary signal.Compared with the traditional method,the new method is proved to improve the overshoot or undershoot and mode mixing problems.(2)For the end effect problem in EMD algorithm,firstly,two solutions to suppress the end effect are analyzed.Then,several typical methods for suppressing the end effect problem are carried on,the thorough research including mirror extension method,waveform matching method and polynomial fitting method and the artificial neural network method,the characteristics of the four typical methods were summarized and compared by simulation.Finally,based on the above research,an inhibition end effect method based on the most similar wavelet is proposed.The validity of the method is proved by the experiments of simulated and actual signals.(3)For the denoising problem in lung signal processing,the types of noise in lung sound signals and the current state of lung noise denoising are summarized.Considering the current single-channel characteristics of collecting lung sounds using electronic stethoscope,a new method ESBSSit is proposed.The new method uses EMD soft threshold denoising combined with single channel blind source separation to eliminate the noise in lung sounds.The method first uses the EMD soft threshold denoising method to eliminate various noises such as muscle friction sound,gastrointestinal motility sound,environment and operation,and then separates the heart sound by using a single channel blind source separation method to obtain a lung signal for eliminating noise.Through the experiments of the three normal lung sounds collected,the effectiveness of the new method is verified from three aspects: waveform,hearing and objective evaluation indicators.At the same time,the SVM-based recognition experiment further verified that the new method can also enhance the lung sound recognition.