This paper studies the variance characteristics of chaotic signal in different conditions and puts forward an adaptive denoising algorithm on account of EMD decomposition layers

by using the Empirical Mode Decomposition (EMD).The arithmetic can adaptively select the IMF layer which needs to be processed

based on the relationship between the maximum variance corresponding layers and the total number of decomposition layers of intrinsic mode function (IMF)

and it also can make intergrated denosing by making use of the lifting wavelet's advantages in the field of updating and predicting.It carried out the experimental study

based on the chaotic background noise from Lorenz and Chen System(adding 10%-100% white gaussian noise) and the measured IPIX radar data.The result shows that:under varying degrees of low noise (30%)

it decreases the error of mean square by at least 30% compared with the methods such as traditional wavelet threshold denoising

and the signal to noise ratio has increased by 1.5db-3.5db

and can effectively reduce the sea clutter noise to increase the detection effect under the background of chaos.