Pulsar navigation is a new type of autonomous navigation. In order to improve its accuracy and real-time, a method of pulsar time of arrival(TOA) estimation combining wavelet transform and compressed sensing was proposed. This method constructed a multi-level redundancy dictionary using wavelet transform for the standard algorithm complexity of this method was lower than the traditional TOA estimation method, and the real-time performance of the algorithm was better as the number of bins in the pulsar signal period increases. This method changed the traditional idea of first denoising and then estimating TOA, which can be embedded in the process of signal denoising, and can be calculated in parallel with the threshold processing of wavelet denoising. The simulation results show that the time consumption of this method is 0.87% and 21.35% of that of the two traditional TOA estimation methods, respectively. And with the shortening of the data duration, the relative advantage of the TOA estimation accuracy of this method is more obvious. This method can not only improve the accuracy of TOA estimation, but also reduce the dependence on the duration of data.