A locally adaptive wavelet de-noising method based on normal inverse Gaussian modal is proposed.Firstly

the db5 wavelet is used to decompose the signal.For those wavelet coefficients which contain a lot of noise

the normal inverse Gaussian modal with good approximation property is constructed as the prior distribution model of those coefficients

on the basis of the model

Bayesian maximum a posteriori estimator is used to estimate the noisy wavelet coefficients and got the realistic wavelet coefficients.Then in the process of posteriori estimation

in order to get the best posteriori approximation model

the particle swarm optimization algorithm is used to select the key coefficient of the model.Finally

new wavelet coefficients are used for the reconstruction of the de-noised signal

and the de-noised signal is gotten.The algorithm is analyzed by simulation and bearing fault signal respectively.Analysis results show that this algorithm has good noise reduction effect

and can efficiently reduce the noise.