Pulse signal random noise reduction method based on AR model spectral estimation

Abstract

The invention relates to a pulse signal random noise reduction method based on AR model spectral estimation, and belongs to the field of information science and medicine integration. The method comprises the steps that firstly, collected pulse signals are divided into two sections of the same length, the power signal to noise ratio of the signals is calculated, a model of random noise signals is built, and the random noise signals are designed; zero filling is performed on the two sections of signals respectively so that the length can be N, Fourier transform is performed on the signals obtained after zero filling, and a phase spectrum is reserved; an AR model is built for the signals obtained after zero filling, and the power spectrum is estimated; zero filling is performed on the noise signals so that the length can be N, and the power spectrum is estimated again; the power spectrum of the noise signals is subtracted from a power spectrum of mixed signals to obtain a power spectrum of effective signals, time domain pulse effective signals are obtained through transformation on the basis of the phase spectrum of mixed signals obtained before noise reduction. According to the pulse signal random noise reduction method, the signal resolution ratio and fidelity are not lowered on the basis that random noise is removed obviously, and the noise reduction effect is good.

Description

technical field [0001] The invention relates to a pulse signal random noise denoising method based on AR model spectrum estimation, and belongs to the technical field of fusion of information science and medicine. Background technique [0002] The various physiological systems in the human body are coupled with each other. The heart and blood circulation system is the most important and comprehensive way to reflect the health status of a person. Therefore, by collecting pulse waves and then analyzing the function of the heart and circulation system, it can reflect the health of the human body more comprehensively. However, the pulse wave signal collected from the human body has a relatively low signal-to-noise ratio, which brings difficulties to the accurate measurement of subsequent parameters, so the removal of noise interference is very important and necessary. [0003] Commonly used pulse signal denoising methods mainly include simple denoising processing in the time do...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a pulse signal random noise denoising method based on AR model spectrum estimation. By estimating the signal-to-noise ratio of pulse signal power, a random noise signal model is established, and the random noise is removed in the frequency domain by using AR model spectrum estimation. The method is used to solve the problem that the useful signal and the noise signal cannot be separated well by traditional time-domain filtering and frequency-domain filtering methods when the useful signal and the noise signal overlap in the frequency domain width; Noise method can produce pseudo-Gibbs phenomenon or introduce the problem of random noise; And in the present invention, the order number of determining AR parameter model is the signal length before zero padding, and the selection of parameter is simple and fixed, but denoising effect is still very good. Ok, it solves the problem of difficulty in parameter selection and the problem of affecting the denoising effect due to improper parameter selection

Images