The conventional timing synchronization methods based on time domain correlation have the problems of timing metric plateau in Additive White Gaussian Noise(AWGN) channel and estimation error in multipath fading channel. To resolve the problems, this paper proposes a novel timing metric using the characteristics of long training symbols in IEEE802.11a and a new timing recovery method based on the new timing metric for Orthogonal Frequency Division MuItiplexing(OFDM)-based WLAN systems. The proposed timing metric is defined as a sum of absolute values of the imaginary parts of all the subcarrier samples. It exhibits a unique characteristic that is very sensitive to the true synchronization point since it has minimum value at the true synchronization point and maximum around the true synchronization point. The simulation results show that the performance of timing synchronization is significantly improved, as a result, the probability of error estimation is lower than 10-4 when Signal-to-Noise Ratio(SNR) i