In this paper, we consider a cognitive radio network with multi-class customer premises equipment (CPE) devices having different data rate requirements. According to the IEEE 802.22 standard, the base station needs to schedule what is called intra-frame sensing process in order to determine the state of the primary users (PUs) channels, where the interval between two consecutive processes is called “sensing interval”. We propose an optimization model based on the partially observable Markov decision process (POMDP) to determine the optimal sensing interval length. Solving this optimization model gives a common sensing interval length for all CPEs in the case of busy channel and assigns to each CPEs its own sensing interval length according to its data rate requirement in the case of idle channel. The numerical results show that the proposed model provides better balancing between approaching the required data rates and protecting the PU compared to the conventional models that apply fixed sensing interval length.