Admission control prevents certain flows from accessing a network with regard to the current utilization level of its resources with the ultimate goal of avoiding congestion and performance collapses, so that, accepted flows receive a sufficient level of Quality of Service (QoS). In this paper, we evaluate the performance of three measurement-based admission control (MBAC) solutions in the context of semantic networks, which autonomously acquire a knowledge on the on-going traffic. Each MBAC solution is carefully parameterized to have identical performance target expressed in terms of maximum tolerable loss rate or either queueing delay. We also include the results that would be obtainable by an ideal admission control so as to benchmark the performance of existing MBAC. An extensive set of simulations, using different methods for representing the background traffic, viz. a Poisson process, a PPBP process and a real trace, were carried out to evaluate the performance of each solution. Our results tend to show that, in case of a target loss rate, when the characteristics of background traffic deviate significantly from a "regular" source (e.g., a Poisson process) to more variable sources (e.g., a PPBP process, a real trace), the "Aggregate Traffic Envelopes" and "Equivalent Capacity" solutions brings satisfactory results, much better than those obtained by the "Measured Sum" solution. However, if one deals with a queuing delay target, the outcomes of any MBAC solution are generally less successful, as they typically tend to deviate further from the ideal admission control, often in an overly conservative way.