Integrated profile, flux and period are three major characteristics of pulsar. Taking advantage of these characteristics, a time domain detecting method is presented based on Bayesian estimation to detect pulsar periodic radiation signals. The signal probability distribution is deduced based on the Poisson distribution model of X-ray pulsar, with noise observation of non-pulse region used as the priori knowledge. The cumulative distribution function of the signal probability is used as a criterion to detect weak signals of X-ray pulsars and extract the phase offset. The RXTE data and the simulation data are used for experimental verification, and the results show that the method outperforms a similar method which is based on the Gauss distribution model, in addition, it can give the phase offset in a certain accuracy.