The coherence of sensing matrix may affect seriously the estimation accuracy of Ultra-Wideband (UWB) channel based on Compressive Sensing(CS), thus a method is proposed which combines the optimized sensing matrix with Bayesian algorithm.Taking the cross cumulative coherence as the principle, the optimized sensing matrix is obtained by optimizing and regularizing the weighted sensing matrix.Bayesian algorithm is then used to reconstruct the UWB channel.Both the theoretical analysis and the experimental results show that the optimized sensing matrix has weaker coherence compared with the sensing matrix, and the estimation accuracy of UWB channel is improved effectively.At the same time, the UWB communication channel can be accurately reconstructed with low compressive ratio and SNR.