This paper investigates the capacity and energy
efficiency of spectrum sharing systems with opportunistic user
selection where a secondary network utilizes spectrum bands
licensed to a primary network under interference regulation. In
spectrum sharing systems, secondary users consume a fraction
of their resources in sensing the channels to the primary users
to comply with the interference constraints. Although more
resources for sensing improve reliability and performance, the
throughput loss due to time overhead and energy loss due to
power overhead should be properly incorporated in performance
evaluation. In this context, we define and derive a new metric –
average capacity normalized by the total energy consumption
– reflecting time and power overhead for spectrum sensing.
Based on the developed framework, the optimal normalizedcapacity
is investigated. We also propose a simple and practical
suboptimal best-n scheme motivated by the infeasibility and
high computational complexity of the optimal strategy, where n
denotes the number of sensing secondary users. Our analytical
and simulation results show that the proposed best-1 scheme is
an energy-efficient technique with near optimality in terms of
the capacity normalized by the energy consumption.