Improved Empirical Mode Decomposition Denoising And Prediction Of Pulse Wave Signals

Interference caused by factors such as pseudo motion and venous noise,high frequency noise and a certain amount of drift occur during the acquisition of the pulse wave.Decomposing the original noisy pulse wave signal into several IMF components using the EMD method.The score function consists of the correlation coefficient value,the Hilbert instantaneous frequency value,and the Hilbert amplitude of each IMF component.In order to avoid the interference of human factors,the concept of the score-to-front ratio is proposed as the dynamic threshold.The dynamic threshold is used to compare the size of the score function to filter the IMF components decomposed by the EMD,so as to achieve the effect of removing noise and drift from the pulse wave.The particle swarm optimization algorithm is used to calculate the optimal score function,which makes the mean squared error root of the pulse wave after denoising the smallest and the signal-to-noise ratio value the largest,that is,the denoising effect is the best.The high-frequency noise and a certain displacement drift are added to the simulated pulse wave in accordance with the noise that may occur in the actually acquired pulse wave.Using the improved EMD method,wavelet analysis method and EMD method proposed in this paper,the simulated pulse wave data of different time length windows are denoised.According to the signal-to-noise ratio and the root mean square error value,the improved EMD method proposed in this paper is obviously superior to the wavelet analysis and EMD method.Using the improved EMD method proposed in this paper,the real-time acquired pulse wave data processing,compared with the denoised pulse wave pattern,can be seen that the high-frequency noise and drift in the original signal have been removed.After removing the noise and drift in the pulse wave by the above steps,the C-C algorithm is used to determine the delay time and the embedded dimension value in the phase space reconstruction to expand the one-dimensional pulse wave into a multi-dimensional pulse wave time series.The support vector machine is used to perform regression prediction on the reconstructed multi-dimensional pulse wave time series to further understand the trend of human pulse wave data.