Cooperative Multiband Spectrum Sensing Using Radio Environment Maps and Neural Networks

. 2023 May 30;23(11):5209. doi: 10.3390/s23115209

Algorithm 1 Central Entity Computed by Traditional Signal Processing
Step 1.1	$The central entity collects the edge detector, binary decision, and power vectors$ $of each PU at determined time instant t from the database . With this$ $information, the central entity reconstructs the occupancy signal$ $O c c u p a t i o n_{T_{t}} \{{R^{'}}_{i} (k)\}$ $and constructs simultaneously, an approximation of the$ $power spectral density P S D_r e c_{T_{t}} \{{R^{'}}_{i} (k)\}$ using only the average powers (power vector) associated to each SU. The length of these frames is set to 1024 samples.
Step 1.2	$Once the reconstruction of the occupation of each SU is performed, the mean$ $value E [O c c u p a t i o n_{T_{t}} \{{R^{'}}_{i} (k)\}]$ is computed.
Step 1.3	$After computing the average value of the Occupation, it is possible to infer the$ $presence of one or several PUs, giving two possible vectors . First, the vector$ $F_{C_v e c t o r} = [F_{C 1}, F_{C 2}, \dots]$ $containing the central frequencies values of detected$ $PUs . The length of this vector will be the number of PUs detected by the$ $algorithm . A \sec ond vector, containing the occupied bandwidths by each$ $detected PU B_{v e c t o r} = [B_{P U_{1}}, B_{P U_{2}}, \dots]$ $, is also obtained . Through the number of$ $\sin gularities and their corresponding widths detected in$ $E [O c c u p a t i o n_{T_{t}} \{R_{i}^{'} (k)\}]$ $, it is possible to estimate how many PUs are in the$ $spectrum and their respective bandwidths (B_{v e c t o r} = [B_{P U_{1}}, B_{P U_{2}}, \dots]$ $) . The$ $F_{C_v e c t o r}$ is integrated by the central frequency values of estimated bandwidth for each detected PU.
Step 1.4	$Knowing now the central frequency and the corresponding bandwidth of each$ $detected PU, it is possible to locate them in each reconstructed PSD$ $P S D_r e c_{T_{t}} \{R_{i}^{'} (k)\}$ $. The mean of the interval B_{P U_{N}}$ $centered on F_{C_{N}}$ $of the$ $P S D_r e c_{T_{t}} \{R_{i}^{'} (k)\}$ $is computed giving the scalar a u x_p s d_{i, N}$ $. In the strict$ $sense, this scalar represents the average power in the carrier, sensed by each$ $SU, where the P U_{N}$ is or should be.
Step 1.5	$With the scalar a u x_p s d_{i, N}$ $of each SU located in a specific coordinate, the REM$ $will be built using a double interpolation . First interpolation is done through$ $the IDW method and for the \sec ond one, the Kriging method is applied . Due to$ $the fact that in our case only eight geographical points are considered (i . e ., only$ $eight SUs are implemented), this double interpolation is carried out with the$ $purpose of having a better precision to describe the behavior of the radio$ $electric space in the environment described later . In this case, each REM is$ $constructed with values a u x_p s d_{i, N}$ $that specifically correspond to the average$ $power of the bandwidth B_{P U_{N}}$ .
Step 1.6	$Finally, the active area of each PU will be determined according to the$ $information collected by each \sec ondary entity S U_{i}$ $. For this, a threshold of$ $- 80$ dBm is used to classify the area estimated by the REMs that was chosen in [41] for a wireless environment. That is, regions in the REM that are above this threshold correspond to the active area of detected PUs.