Probability Distribution Mining of Time-Sensitive Spectrum Sensing Data

In the multi-dimensional spectrum space, the spectrum has strong time-varying characteristics. In this way, it is of great significance to estimate the changes of spectral noise and channel quality over time. The traditional Spectrum sensing technology as a key component in cognitive radio technology groups has make significantly contribution to improve the wireless spectrum utilization efficiency. However, many are not enough to judge whether the channel is available. The judge problem can be improved by estimating the probability distribution of spectrum sensing data changes. In this paper, we focus on spectrum sensing data and channel sensing data which are collected over months in various terrain environments. We observe that all data are time sensitive and may conform to some probability distribution. By a series of hypothesis testing on these data, we found that 77% of the samples with spectral noise changing data obey the exponential distribution, and that most channel quality changing data are basically subject to exponential distribution. This result can help us to predict and select spectrum holes with sound confidence.