Time and Frequency Varying Noise Floor Estimation for Spectrum Usage Measurement

We investigate noise floor (NF) estimation for FFT-ED (Energy Detection based on FFT)-based spectrum usage measurement in the context of smart spectrum access (SSA), in which spectrum usage information of primary users (PUs), such as channel occupancy rate (COR), will be exploited by secondary users (SUs). In FFT-ED, the NF has to be estimated to set a decision threshold for ED appropriately. In general, the NF is frequency-dependent and its level changes with time leading to the need of estimating the NF regularly while performing the spectrum usage measurement. In this paper, we propose an NF estimation method which exploits prior information regarding the shape of NF and forward consecutive mean excision (FCME) algorithm. Numerical and experimental evaluations show the proposed method enables an accurate NF estimation considering the time and frequency dependencies of the NF. Moreover, we show the proposed method can obtain the almost desired detection performance, but can not the comparative method (the original FCME method).