SUMMARYThe handover process was widely studied in the literature in the past decade. While significant efforts addressed the mean handover rate, few attempted to characterise the distribution of the number of handovers under the assumption that the call and the cell residence times (CRTs) are generally distributed. Motivated by the increasing impact of handovers on the design and performance of the control and data planes in the emerging cellular networks, in this paper, we propose a new analytical framework to obtain the probability generating function (PGF) of the number of handovers. Our work covers three generic session behaviours, including new, roaming and location triggered calls, and thus facilitates the calculation of the mean number of handovers for such session types. When the distributions have rational Laplace transforms, we obtain closed form solutions that are simple-to-compute and hence provide cellular network designers and researchers with deeper insights into the performance of mobility aware wireless protocols for different application scenarios. Our results show that handover models for new sessions can reasonably approximate location triggered sessions while separate treatment for roaming sessions is generally required. The numerical examples based on the generic hyper-Erlang session durations show the practicability of our framework.