This paper investigates a signal area (SA) estimation method for wideband and long time duration spectrum measurements for dynamic spectrum access. SA denotes the area (in time/frequency domain) occupied by the primary user's signal. The traditional approach, which utilizes only Fourier transform (FT) and energy detector (ED) for SA estimation, can achieve low complexity, but its estimation performance is not very high. Against this issue, we apply post-processing to improve the performance of the FT-based ED. Our proposed method, simple SA (S-SA) estimation, exploits the correlation of the spectrum states among the neighboring tiles and the fact that SA typically has a rectangular shape to estimate SA with high accuracy and relatively low complexity compared to a conventional method, contour tracing SA (CT-SA) estimation. Numerical results will show that the S-SA estimation method can achieve better detection performance. The SA estimation and processing can reduce the number of bits needed to store/transmit the observed information compared to the FT-based ED. Thus, in addition to improved detection performance it also compresses the data.

Abstract-Two-layer smart spectrum access (SSA) consists of spectrum sharing based on dynamic spectrum access (DSA: first layer) and spectrum awareness system (SAS: second layer). A main role of SAS is providing useful statistical information in terms of spectrum usage by long term, wide-band and widearea measurements. In this paper, we focus on signal area (SA) estimation which is a core signal processing in SAS to understand the spectrum usage. Specifically, SA estimation is used as post processing for energy detector outputs. It has been shown that Simple-SA (S-SA) estimation can enhance the spectrum usage measurement performance, but it inherently increases false alarms. For this issue, efficient false alarm cancellation technique, L-shaped false alarm cancellation (L-FC), is proposed in this paper. Numerical evaluations show that the proposed method can achieve proper detection performance while the computational cost is small compared to other methods.