Histogram partitioning algorithms for adaptive and autonomous threshold estimation in cognitive radio–based industrial wireless sensor networks

Modern energy detectors typically use adaptive threshold estimation algorithms to improve signal detection in cognitive radio–based industrial wireless sensor networks (CR-IWSNs). However, a number of adaptive threshold estimation algorithms often perform poorly under noise uncertainty conditions since they are typically unable to auto-adapt their parameter values per changing spectra conditions. Consequently, in this paper, we have developed two new algorithms to accurately and autonomously estimate threshold values in CR-IWSNs under dynamic spectra conditions. The first algorithm is a parametric-based technique termed the histogram partitioning algorithm, whereas the second algorithm is a fully autonomous variant termed the mean-based histogram partitioning algorithm. We have evaluated and compared both algorithms with some well-known methods under different CR sensing conditions. Our findings indicate that both algorithms maintained over 90% probability of detection in both narrow and wideband sensing conditions and less than 10% probability of false alarm under noise-only conditions. Both algorithms are quick and highly scalable with a time complexity of O(V), where V is the total number of input samples. The simplicity, effectiveness, and viability of both algorithms make them typically suited for use in CR-IWSN applications.