In this paper, timing synchronization is investigated for a decode-and-forward (DF) cooperative communication system with a single source, a single destination, and multiple relays. In the multiple access phase where multiple relays simultaneously transmit to the destination, timing to be carried out by the destination receiver involves estimating multiple delay parameters associated with different relays. The existing maximum-likelihood (ML) multiple delay estimator needs exhaustive search over the estimation range, and the complexity of the ML estimator exponentially increases as the number of relays or the resolution increases. We consider a correlation-based timing estimator to save computational complexity. The proposed method requires that each relay transmits a pseudorandom noise (PN) sequence as the training symbols for synchronization. At the destination, the discrete superimposed signal is first interpolated to the required resolution and then correlated with each relay's PN sequence modulated waveform template. The peak of each correlation yields the estimation of the delay for the corresponding relay. The correlation output at the peak also provides an estimation of the fading channel, which can be seen as a by-product of the timing estimator. The proposed timing estimator saves substantial complexity compared with the ML estimator and is able to achieve satisfactory performance, as demonstrated by the simulation results.