Abstract The problem of calculating the local and global decision thresholds in hard decisions based cooperative spectrum sensing is well known for its mathematical intractability. Previous work relied on simple suboptimal counting rules for decision fusion in order to avoid the exhaustive numerical search required for obtaining the optimal thresholds. However, these simple rules are not globally optimal as they do not maximize the overall global detection probability by jointly selecting local and global thresholds. In this paper, a globally optimal decision fusion rule for Primary User signal detection based on the Neyman-Pearson (NP) criterion is derived. The algorithm is based on a novel representation for the global performance metrics in terms of the regularized incomplete beta function. Based on this mathematical representation, it is shown that the globally optimal NP hard decision fusion test can be put in the form of a conventional one dimensional convex optimization problem. A binary search for the global threshold can be applied yielding a complexity of $O(\log_2(N))$, where $N$ represents the number of cooperating users. The proposed optimal scheme outperforms conventional counting rules, such as the OR, AND, and MAJORITY rules, offering significant SNR gain in dense cognitive radio networks.
Introduction
In Cognitive Radio (CR) networks, unlicensed "secondary users" (SUs) must detect if a licensed "primary user" (PU) is active to avoid interference. Due to fading and shadowing—the "hidden terminal problem"—a single SU might miss the PU. To mitigate this, SUs cooperate by sharing sensing results with a Fusion Center.
There are two primary methods for combining these results: Soft Decision (SD), which shares raw energy data, and Hard Decision (HD), where each SU sends a 1-bit "yes/no" vote. While HD is bandwidth-efficient, it is notoriously difficult to optimize. Finding the "globally optimal" pair of thresholds—a local one for each SU and a global one for the Fusion Center—historically required exhaustive numerical searches. Consequently, researchers often settled for suboptimal "counting rules" like OR or MAJORITY.
Methodology
This work derives a globally optimal fusion rule based on the Neyman-Pearson (NP) criterion, maximizing the probability of detection ($Q_d$) for a fixed allowable probability of false alarm ($\alpha$). The key technical innovations include:
- Beta Function Representation: Global performance metrics, which follow a binomial distribution, are expressed using the regularized incomplete beta function.
- Convex Optimization: The authors prove that the global detection probability is log-concave with respect to the global threshold. This transforms the search into a conventional 1D convex optimization problem.
- Binary Search Algorithm: Utilizing the convex nature of the problem, a binary search finds the optimal global threshold ($n$). Once $n$ is determined, the local threshold ($\lambda$) is calculated via a closed-form inverse function.
The algorithm achieves a complexity of $O(\log_2 N)$, making it highly scalable for dense networks with hundreds or thousands of sensors.
Results and Discussion
Numerical simulations using Rayleigh fading and AWGN models validate the analytical derivation. Key findings include:
- Performance Gains: In a network of $N=16$ users, the proposed optimal rule provides a 1 dB SNR gain over the OR rule and a 1.5 dB gain over the MAJORITY rule.
- Dense Network Advantage: The SNR gains become more pronounced as the number of cooperating users increases. While the OR rule is often near-optimal for small networks, it fails as the network becomes dense, where the optimal threshold shifts toward the middle.
- Sensing Time Trade-off: Increasing the number of sensing samples ($M$) from 6 to 24 can offer up to a 4 dB SNR gain, though this increases sensing delay.
Conclusion
By representing performance metrics through the regularized incomplete beta function, this paper solves the mathematical intractability of optimizing hard-decision spectrum sensing. The resulting $O(\log_2 N)$ algorithm allows for real-time threshold recalculation in practical, dense cognitive radio deployments. Future work may address non-ideal reporting channels and heterogeneous networks where SUs experience varying average SNR levels.
Summary updated