Abstract: In this paper, an efficient hybrid sampling algorithm introducing the random walk method into the reversible jump Markov Chain Monte Carlo (RJMCMC) method is developed to jointly detect and estimate the wideband signals impinging on passive sensors, which combines the local sampling and the whole space sampling, and can get random number more obeying to target distribution in a shorter time. The method is proposed on the basis of Bayesian theory, and the statistical characteristic of the noise and signals. The accuracy and the validity of this method are well verified by the simulation.