The use of artificial neural networks (ANNs) for transit time estimation is investigated. ANNs are proposed as an alternative to widely employed traditional techniques such as cross correlation and the cross spectrum, which are sensitive to the presence of noise and require a large volume of data for their calculation. The ANN employed is based on interactive activation and competition and has been found able to correctly estimate the current transit time from short records of signals generated by simulation, quickly follow changes in transit time, and detect when the transit time falls outside a predefined expected range. By appending a backpropagation ANN, the on-line estimation of decimated transit times is also allowed.