The Research Of Denoising Based On EMD And HMT

In this paper,the main work can be summarized as two partments: Remove the baseline drift in electrocardiogram and Denoising images.(1)Baseline drift correction method is a kind of traditional testing and mean filtering detection. Such as spline function method, FIR and IIR filter method, adaptive filter design method. In recent years, the wavelet analysis method to be heard, and also the correction of baseline drift Mallat wavelet algorithm. In this paper, we proposed EMD decomposition method and the wavelet transformation based on EMD method. EMD decomposition is a decomposition of non-stationary signals, it will break down complex signals into a number of intrinsic mode functions(IMF) by order of the frequency of high and low , each IMF is a single-component signal. The differences between it and the wavelet method are its posteriori and do not need to pre-selected basis functions.But according to the characteristics of the signal , can generate the appropriate adaptive mode function, the mode function can be a very good signal at any time to reflect local characteristics of the frequency.EMD decomposition method:Therefore, the EMD method is removing a number of low-frequency component of the IMF and r(n), and then reconstruction the signal. The wavelet-based based on EMD method is superior to the stability.The wavelet-based based on EMD method:Do as following:1)x(t)do EMD decomposition,obtain IMFs;2)x(t)subtract any IMFs,obtain R_n(t).3)Choose the appropriate wavelet function and decomposition level n fo, wavelet transform with R_n(t),obtain approximate cA_n(t). 4)x(t)subtract cA_n(t)obtain the denoising signal x'(t).x'(t) = x(t) - cA_n(t) (5.5)x' (t) is the signal after denoising; cA_n(t) is the baseline drift that have denoised. EMD decomposition method and wavelet-based EMD method is basically to ensure that no loss of ECG signal . It can retain additional information signal effectively and suppress the baseline drift noise, and also for a long period of ECG records ,it can be made to deal with the same good results. In particular, the wavelet method based on EMD can be better to retain the characteristics of the signal, better accurate, more significant effects .(2) This paper also discussed the denoising of Gaussian noise images. the hidden Markov tree(HMT) models use a probabilistic tree to model Markovian dependencies. The hidden states capture the dependencies between the wavelet coefficients,and HMT is useful tool for image processing. In this paper, we extend the HMT modeling framework to the complex wavelet transform.At the same time discussed the Dual-Tree Complex Wavelet Transform Construction Principle and the nature of Dual-Tree Complex Wavelet Transform .It is not only inherited the tradition of the advantages of wavelet transform, but also with the approximate translation invariance, multi-directional, limited redundancy and efficient computing. Dual-Tree Wavelet eliminate the Gibbs phenomenon arising on image denoising. We proposed Dual-Tree Wavelet-domain HMT model algorithm for image denoising .The HMT model in the DT-CWT domain for image denoising do as following:(1)X' = X + n,do DT-CWT inX' ,obtain complex wavelet coefficients (w)i. (2)Decomposing (w)_i into real and imaginary parts,(w)_i = (a)_i + j * (b)_i (3)Obtain (a)'_i andj * (b)'_i after the HMT denoising with(a)_i and(b)_i(5)(w)'_i do Dual-Tree Complex Wavelet inverse transform and Obtain Z.(?) : the HMT model parameters.(3)Finally, we verify our proposed algorithms used the actual ECG data and lena image ,and compare the test results.