The core problem of the brain-computer interface (BCI) based on neural signal is estimating neural firing rate from a spike train and then using neural population decoding algorithm to decode movement trajectory.In this artical, we review the theoretical basis of both classic and current firing rate estimations and compare the advantages and drawbacks of these methods. At the same time we also review the decoding algorithm which using neural firing rate to decode movement trajectory in brain- computer interface: population vector algorithm, linear filter and kalman filter. At last, some results applying these estimators of firing rate to decode arm movement in BCI are introduced. The results show apparently different performance of the different firing rate estimators, while minimal differences are observed in the actual application of BCI.