Skip to Main Content
This paper introduces a hybrid energy detection approach to spectrum sensing in cognitive radio. Conventional energy detection technique (with N samples) is based on two hypotheses - i) where all N are signal samples corrupted by noise, ii) where all N are noise-only samples. Based on these two hypotheses a threshold is set for declaring a given N-sample signal as belonging to either of the two cases. This paper considers those cases which occur in practice - M samples out of N being signal corrupted with noise and the rest being noise-only, 0 ≤ M ≤ N, thus covering the conventional two hypotheses as well. The method described here, referred to as “hybrid energy detection”, proposes a combination of a 32-sample detector with a 16-sample detector which yields an improvement in performance in most such mixed signal-noise cases. The “hybrid energy detection” method requires certain thresholds to be set, in order to maximize the gain. We use simulated annealing to find those threshold values. The cost function used for simulated annealing takes into consideration the ratio of sensing time interval to transmit time interval in the secondary user.