Abstract: Pulsar time defined by single pulsar is influenced by several noise resources, to weaken these influence for gaining a more stable time scale, one can take ensemble analysis method to obtain ensemble pulsar time. Three algorithms of ensemble pulsar time are presented: classical weighted average algorithm, to gain the best long-term stability of ensemble pulsar time, one can choose the weight according to the long-term stability of each pulsars; the residuals of single pulsar can be decomposed by wavelet analysis in wavelet domain, and the component of different frequency range can be obtained. Then we can apply wavelet analysis algorithm to integrate the pulsar time, the weights are chosen according to reciprocal of wavelet average square error; The pulsar timing residuals are caused by reference atomic clock and pulsar itself, Wiener filtration analysis algorithm allows the separation of the contributions of an atomic clock and a pulsar itself to the post-fit pulsar timing residuals. The method allows to filter the atomic scale component from the pulsar phase variations. These three algorithms have been applied to the timing data of the millisecond pulsars PSR B1855+09 and PSR B1937+21. The result has indicated that the ensemble pulsar time obtained after the ensemble algorithm to pulsar time defined by several pulsar time weakens the influence of noise on a great extent, consequently improves the stability, furthermore the stability of the ensemble pulsar time obtained by wavelet analysis algorithm and Wiener filtration are better than the stability of the ensemble pulsar time obtained by sample weighted average algorithm. Obviously wavelet analysis algorithm and Wiener filtration are more feasible approach to deal with time-frequency signals.

Key words: ensemble pulsar time；wavelet analysis；Wiener filtration