The wavelet transform is introduced on the basis of UPF, assuming that the system noise is Gaussian distribution, and the parameters of the target state, such as position, velocity, etc., are in the low frequency band relative to the system noise, and the importance function is selected as the transfer prior. At this time The particles sampled from the transfer a priori are multi-decomposed by wavelet transform to obtain signal components of different frequency bands, and then the high-frequency components of the reaction noise are removed, and the remaining low-frequency signals are used for wavelet reconstruction to obtain the r part of the noise. For particles, the new transition prior of these reconstructed particle reactions has a smaller variance than the original transition prior, so that a sharper probability density function can be obtained, thus reducing the variance of importance weights.
正在翻译中..