One-decoy-state quantum key distribution with infinite length pulses can even approach the asymptotic case, i.e., infinite decoy states and infinite length pulses. Due to the limitation of experimental time, we must consider the statistical fluctuation caused by the finite number of pulses. The central limit theorem analysis method is an effective method, which can easily give the upper and lower bounds of experimental observations. Combining the central limit theorem method with the theorem that the ratio of two independent normal random variables can be approximated as a normal random variable under certain conditions, we propose a more efficient method to estimate the secure key rate. Compared with using central limit theorem method directly, our method has fewer bounds, smaller estimation deviation and higher estimation accuracy. The simulation results show that our method can significantly improve the secure key rate and transmission distance when considering statistical fluctuation and keeping the pulse number and security parameters unchanged, especially in the case of small pulse number (e.g. 107).