LAUSR

In recent years, blind deconvolution has been widely used in machinery condition monitoring and has been proven to be an effective method for impact detection. Considering that the ultrawideband (UWB) measuring system adopts Gaussian impulses as the transmitting signal, the improved deconvolution algorithm explained in this article extracts the UWB impulse series from the mixture of real signal, noise, and jamming signals. Then, the ability of entropy to characterize the impulse is investigated to improve the deficiency of traditional methods, which depend on kurtosis. Last but not least, using approximate negative entropy as the sparsity measure function of the deconvolution, the sensitivity of kurtosis is avoided, and a fast deconvolution algorithm is proposed on this basis. Experiments show that the negative entropy deconvolution can better recover the UWB impulse characteristics under strongly suppressive jamming. The root-mean-square error (RMSE) between the source and the recovered signal is less than 0.05 even when the average signal-to-jamming radio (ASJR) is below −30 dB. Moreover, iterations can be reduced to 1/3 or lower compared with the classic deconvolution algorithm.