Weak feature extraction based on zerophase unsampling wavelet transform

In order to avoid the influence on decomposition result exerted by non-linear phase which was conduced by the non-symmetry of orthogonal wavelet, the zero phase undecimated discrete wavelet transform, which was undecimated discrete wavelet decompose algorithm with zero phase characteristic, was presented. For the sake of reducing the amount of computation, the quick convolution algorithm based on overlap preserving method and circumferential convolution replaced the traditional convolution algorithm in zero phase undecimated discrete wavelet transform. The transform not only overcomes the many defects (such as overlap of frequency and shift variance) in Mallat algorithm, which are caused by down sampling, but also eliminates the displacement and distortion phenomenon in decomposition results of conventional wavelet transform. So the zero phase undecimated discrete wavelet transform is very suitable for weak feature extraction. The weak feature extraction example analysis shows that the zero phase undecimated discrete wavelet transform can effectively extract the weak feature and has an advantage over others. © 2020, Editorial Board of Journal of Huazhong University of Science and Technology. All right reserved.